NLP_basics.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# NLP for Health and Social Care 2018\n",
    "## Skills Development Workshop\n",
    "-----------------------------------------\n",
    "\n",
    "\n",
    "\n",
    "### Before we can do anything we need data!\n",
    "\n",
    "The data that we are interested in are held in pdf file formats, which is typical for published reports.\n",
    "\n",
    "We've manually stored a sample of pdf's in the repo, the links for the source of these files can be found in the issues. \n",
    "\n",
    "We have not included a function for webscraping pdfs's in this notebook, this is something that could be added to enable automatic collection of large numbers of documents. (tip: scrape documents into file structure as in repo to enable seemless integration with the code below) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### read pdf file and convert to text\n",
    "\n",
    "let's start off by importing a single pdf file - sample.pdf - and convert into a text file\n",
    "\n",
    "to this we make use of pdfminer3k libraries (here for more info https://github.com/jaepil/pdfminer3k )\n",
    "(if using for the first time you will need to load into anaconda using conda install syntax below in a terminal window)\n",
    "\n",
    "this code creates a function called convert, which can be called with the file path for the pdf file\n",
    "\n",
    "install pdfminer3k if using for first time (paste syntax below in Terminal window)\n",
    "\n",
    "```\n",
    "conda install -c conda-forge pdfminer3k Python libraries\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Fremantle Trust\n",
      "Apthorp Care Centre\n",
      "\n",
      "Inspection report\n",
      "\n",
      "Nurserymans Road\n",
      "London\n",
      "N11 1EQ\n",
      "\n",
      "Tel: 02082114000\n",
      "Website: www.fremantletrust.org\n",
      "\n",
      "Ratings\n",
      "\n",
      "Date of inspection visit:\n",
      "09 October 2017\n",
      "10 October 2017\n",
      "11 October 2017\n",
      "\n",
      "Date of publication:\n",
      "04 January 2018\n",
      "\n",
      "Overall rating for this service\n",
      "\n",
      "Requires Improvement  \n",
      "\n",
      "Is the service safe?\n",
      "\n",
      "Is the service effective?\n",
      "\n",
      "Is the service caring?\n",
      "\n",
      "Is the service responsive?\n",
      "\n",
      "Is the service well-led?\n",
      "\n",
      "Requires Improvement     \n",
      "\n",
      "Requires Improvement     \n",
      "\n",
      "Good     \n",
      "\n",
      "Good     \n",
      "\n",
      "Requires Improvement     \n",
      "\n",
      "1 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "Summary of findings\n",
      "\n",
      "Overall summary\n",
      "\n",
      "This inspection took place on 9, 10 and 11 October 2017 and was unannounced. During the last inspection \n",
      "on 1 February 2017 we found the service was in breach of four legal requirements and regulations \n",
      "associated with the Health and Social Care Act 2008. We found that people who used the service were not \n",
      "protected against the risks associated with their care and treatment. Medicines were not managed safely. \n",
      "There were not always adequate staffing levels at meal times. Staff were not receiving consistent support \n",
      "through regular supervision. Not all people were supported to maintain their personal dignity. \n",
      "\n",
      "At a previous inspection in July 2016, we found the service to be inadequate overall. We placed Apthorp Care\n",
      "Centre into special measures and imposed a condition on the provider's registration to submit monthly \n",
      "audit reports to the Care Quality Commission (CQC). Following the inspection in February 2017, although \n",
      "some improvement had been made, the service remained in special measures and the conditions on the \n",
      "provider's registration remained in place. \n",
      "\n",
      "This inspection was carried out within the six-month time frame to check if improvements to the quality of \n",
      "care had been implemented. At this comprehensive inspection we found the registered provider had taken \n",
      "action to achieve compliance with some of the regulations previously identified as non-compliant during \n",
      "the comprehensive inspection in February 2017. However, we found repeated breaches of Regulations 17 \n",
      "and 18 of the Health and Social Care Act 2008 (Regulated Activities) Regulations. \n",
      "\n",
      "Apthorp Care Centre is a purpose built residential care home that is registered to provide accommodation \n",
      "for up to 108 people who require personal care. The home is split into three floors that contain units called \n",
      "'flats'. At the time of our inspection there were 75 older people living at the home, many of whom were living \n",
      "with dementia. \n",
      "\n",
      "There was a registered manager in place. A registered manager is a person who has registered with the Care \n",
      "Quality Commission to manage the service. Like registered providers, they are 'registered persons'. \n",
      "Registered persons have legal responsibility for meeting the requirements in the Health and Social Care Act \n",
      "2008 and associated Regulations about how the service is run.\n",
      "\n",
      "At our last inspection we identified that staff were not effectively deployed around mealtimes. At this \n",
      "inspection we found that although there were sufficient staff on duty, at key times throughout the day such \n",
      "as mealtimes, we found that staff were not effectively deployed to ensure people received support with \n",
      "eating in a timely manner. \n",
      "\n",
      "At our last inspection, we found that appropriate fluid monitoring records were not kept for people who \n",
      "were at risk of dehydration. At this inspection we identified repeated concerns in this area. \n",
      "\n",
      "At this inspection, we found that although risk assessments were now identifying risks individual to people \n",
      "and staff were receiving appropriate guidance, some risks to people were, in practice, not actively mitigated \n",
      "\n",
      "2 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "against. \n",
      "\n",
      "Improved systems and processes were in place to monitor quality of care. However, quality monitoring \n",
      "required further implementation and embedding. \n",
      "\n",
      "Medicines were now safely stored and administered. Appropriate records were kept of medicine \n",
      "administrations and regular medicines audits took place. \n",
      "\n",
      "Staff training, supervisions and appraisals were monitored and updated regularly. Systems had been \n",
      "implemented to ensure a better oversight of when staff training, supervisions and appraisals were due. \n",
      "\n",
      "Procedures and policies relating to safeguarding people from harm were in place and accessible to staff. All \n",
      "staff had completed training in safeguarding adults and demonstrated an understanding of the types of \n",
      "abuse to look out for and how to raise safeguarding concerns.\n",
      "\n",
      "Care plans were person centred and reflected what was important to the person. Care plans provided \n",
      "appropriate guidance to enable staff to deliver person centred care in line with people's preferences.\n",
      "\n",
      "Accidents and incidents were monitored and analysed on a monthly basis. \n",
      "\n",
      "People engaged in a comprehensive activities programme delivered by dedicated activities co-ordinators. \n",
      "\n",
      "People and relatives spoke positively of the caring nature of care staff. \n",
      "\n",
      "The home was clean and well-maintained; however we identified a concern regarding water temperatures. \n",
      "The registered provider took immediate action to resolve the situation. \n",
      "\n",
      "Consent to care was obtained from the appropriate person. Care plans specifying best interest needs were \n",
      "in place. Staff had received training on the Mental Capacity Act 2005 (MCA) and staff understood the \n",
      "importance of obtaining consent from people. Deprivation of Liberty Safeguards (DoLS) were applied for \n",
      "and monitored. \n",
      "\n",
      "At this inspection we identified breaches of regulations. You can see what action we have asked the provider\n",
      "to take at the back of the full version the report. However, as the provider had demonstrated significant \n",
      "improvements, the service is no longer rated as inadequate for any of the five questions and therefore no \n",
      "longer remains in special measures.\n",
      "\n",
      "3 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "The five questions we ask about services and what we found\n",
      "\n",
      "We always ask the following five questions of services.\n",
      "\n",
      "Is the service safe?\n",
      "\n",
      "Requires Improvement  \n",
      "\n",
      "The service was not always safe. Staff were not effectively \n",
      "deployed at certain times of the day to ensure people's needs \n",
      "were met. \n",
      "\n",
      "Medicines were managed safely.\n",
      "\n",
      "Individual risks were assessed and guidance was provided to \n",
      "staff, however we found instances were assessed risks, in \n",
      "practice, were not mitigated. \n",
      "\n",
      "People told us they felt safe. Staff were aware of different types of\n",
      "abuse and what steps they would take if they had any \n",
      "safeguarding concerns.\n",
      "\n",
      "Recruitment procedures were in place to ensure staff members \n",
      "were fit to undertake their roles.\n",
      "\n",
      "Is the service effective?\n",
      "\n",
      "Requires Improvement  \n",
      "\n",
      "The service was not always effective. The service was not \n",
      "completing health monitoring charts effectively to ensure people\n",
      "at risk of dehydration received sufficient hydration. \n",
      "\n",
      "Staff had access to regular training, supervisions and appraisals \n",
      "which supported them to carry out their role.\n",
      "\n",
      "Mental capacity and Deprivation of Liberty Safeguards were \n",
      "understood and principles of the code of practice were being \n",
      "followed.\n",
      "\n",
      "People were given the assistance they required to access \n",
      "healthcare services and maintain good health.\n",
      "\n",
      "Is the service caring?\n",
      "\n",
      "The service was caring. There were positive relationships \n",
      "between staff and people using the service. Staff treated people \n",
      "with respect and dignity.\n",
      "\n",
      "Staff had a good knowledge and understanding on people's \n",
      "\n",
      "4 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "Good  \n",
      "\n",
      "\f",
      "background and preferences.\n",
      "\n",
      "People were supported to maintain independence, where \n",
      "possible.\n",
      "\n",
      "Is the service responsive?\n",
      "\n",
      "The service was responsive. Care plans were person centred.\n",
      "\n",
      "People had access to a variety of activities. \n",
      "\n",
      "The home had a complaints policy in place; complaints were \n",
      "investigated and responded to. People and relatives knew how \n",
      "to complain if they needed to.\n",
      "\n",
      "Good  \n",
      "\n",
      "Is the service well-led?\n",
      "\n",
      "Requires Improvement  \n",
      "\n",
      "The service was not always well-led. We identified repeated \n",
      "concerns at this inspection. \n",
      "\n",
      "Quality assurance systems were in place, however further \n",
      "embedding was required. Negative feedback received from \n",
      "people in a quality survey was not acted upon in a timely \n",
      "manner. \n",
      "\n",
      "We received positive feedback regarding the registered manager \n",
      "from people, relatives and staff. People and relatives told us they \n",
      "had seen the quality of care improve at Apthorp Care Centre.\n",
      "\n",
      "5 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " \n",
      "\n",
      "Apthorp Care Centre\n",
      "\n",
      "Detailed findings\n",
      "\n",
      "Background to this inspection\n",
      "We carried out this inspection under Section 60 of the Health and Social Care Act 2008 as part of our \n",
      "regulatory functions. This inspection checked whether the provider is meeting the legal requirements and \n",
      "regulations associated with the Health and Social Care Act 2008, to look at the overall quality of the service, \n",
      "and to provide a rating for the service under the Care Act 2014.\n",
      "\n",
      "This inspection took place on 9, 10 and 11 October 2017 and was unannounced. The inspection team \n",
      "consisted of a lead inspector, three supporting inspectors, a pharmacist inspector, a specialist nurse advisor \n",
      "and two experts by experience. An expert by experience is a person who has personal experience of using or \n",
      "caring for someone who uses this type of care service.\n",
      "\n",
      "Before the inspection we reviewed the information we held about the service such as statutory notifications \n",
      "and safeguarding alerts. We also looked at the action plan that the service had provided to the CQC \n",
      "following the last inspection and the monthly updates received. We obtained feedback from the local \n",
      "authority quality monitoring team and one involved health professional. \n",
      "\n",
      "During the inspection we spoke to 30 people who used the service, 10 relatives, 20 care workers, \n",
      "housekeeping and maintenance staff, four activities staff, three floor managers, two deputy managers, the \n",
      "registered manager and two regional directors. We obtained feedback from two visiting health and social \n",
      "care professionals. \n",
      "\n",
      "We also used the Short Observational Framework for Inspection (SOFI). SOFI is a way of observing care to \n",
      "help us understand the experience of people when they may not be able to tell us themselves.\n",
      "\n",
      "We looked at 14 people's care plans and risk assessments, 14 Medicines Administration Records (MAR), 13 \n",
      "staff files and records related to the management of the service such as quality assurance documents, \n",
      "health and safety records and the training matrix.\n",
      "\n",
      "6 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " Is the service safe?\n",
      "\n",
      "Requires Improvement\n",
      "\n",
      "Our findings  \n",
      "At our last inspection on 1 February 2017, we found that staffing levels were not sufficient around meal times\n",
      "to ensure that people were supported to eat and drink. We found at this inspection that although some \n",
      "improvements had been made and that there were sufficient staff available in the home overall to safely \n",
      "ensure people's care needs were met, we remained concerned about how staff were deployed to ensure \n",
      "people received support in a timely manner during meal times. \n",
      "\n",
      "We spent time looking at the staffing arrangements in place to support people living at Apthorp Care Centre.\n",
      "We spoke with people who used the service, relatives, and staff, looked at staffing rotas, the dependency \n",
      "tool and observed the support offered throughout the day.\n",
      "\n",
      "We observed lunch time across three days in all areas of the home. We found that in areas where people \n",
      "needed additional support to eat or required staff to prompt them, this was not always happening. We \n",
      "observed that one person waited in excess of thirty minutes to be supported by care staff after their lunch \n",
      "had been served, although it was noted that the staff member apologised to the person for the delay when \n",
      "they started to assist them to eat. We also observed that in one flat, two care staff were supporting five \n",
      "people to eat, two of whom were entirely dependent on staff for support with eating. Three people required \n",
      "prompting and encouragement to eat and were refusing their meals. Both staff members attempted to \n",
      "support all five people; however, it meant that people were receiving their meals in a stop-start fashion and \n",
      "food was left to get cold. We observed some senior staff members assist care staff during lunchtime; \n",
      "however, we received consistent feedback from staff and a professional involved with the service that senior \n",
      "staff did not usually assist at mealtimes.\n",
      "\n",
      "Since the last inspection, we found that the service no longer relied on agency staff to ensure staffing levels \n",
      "were at a safe level. The regional director told us that this had been a significant achievement in stabilising \n",
      "the workforce and ensuring that the quality of care improved. Rotas confirmed that 21 care staff were on \n",
      "duty in the early part of the day, 20 care staff in the afternoon and evenings with nine care staff on duty at \n",
      "night which was reflected in the provider's dependency tool. We received a mixed response from people \n",
      "who used the service when asked if they thought there were enough staff. Comments received from people \n",
      "included, \"Sometimes a bit short. Not a big thing. Carers come and chat and talk when doing something\", \n",
      "\"Yes I do think there is enough\", \"No it is woeful. There is not enough staff here\" and \"Staff very helpful and \n",
      "enough of them.\" Relatives told us, \"I noticed recently new staff recruited i.e. floor managers and they take \n",
      "notice\" and \"They ring me and keep me updated and you get continuity of staff. All good now. Staff were \n",
      "really poor and there was bad treatment; none of it now.\" \n",
      "\n",
      "We discussed our concerns with the home management who advised that they recognised staffing at \n",
      "mealtimes was a challenge and following the inspection, submitted an action plan detailing that mealtimes \n",
      "would be protected and all staff in the home including admin and domestic staff, with appropriate training, \n",
      "would be drafted in to support with mealtimes. \n",
      "\n",
      "There were not always sufficient numbers of staff suitably deployed at the service to ensure people's needs \n",
      "\n",
      "7 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "were met. This was a continued breach of Regulation 18 of the Health and Social Care Act 2008 (Regulated \n",
      "Activities) Regulations 2014.\n",
      "\n",
      "At our last inspection in February 2017, we found that although risk assessments were identifying most risks \n",
      "associated with people's care and treatment, there were instances were risks had not been considered \n",
      "appropriately. At this inspection, we found that the provider had mostly addressed the issue. \n",
      "\n",
      "Risk assessments were person centred, up to date and reviewed regularly. They covered risks to people such\n",
      "as skin integrity, moving and handling, falls, risks associated with the use of bedrails, smoking, alcohol \n",
      "consumption and health conditions such as diabetes, mental health conditions and associated behaviours. \n",
      "Staff were able to explain how they would work with people's identified risks. For example, one person's \n",
      "care record detailed that the person had been identified as being at risk of falls. Staff supported this person \n",
      "regularly to ensure the person continued to maintain their independence and felt safe. The person told us, \n",
      "\"The staff checks on me during the day and at night and accompanies me to the toilet. At home I used to \n",
      "have small accidents. Not here.\" \n",
      "\n",
      "Another person's care record indicated that they were at risk of developing pressure ulcers. Their care record\n",
      "specified they should be supported to turn over in bed to relieve pressure on their skin. They were supported\n",
      "to do this every two hours. Staff had signed a chart to confirm that they had done this. There were \n",
      "repositioning charts for the past ten days confirming that they had done so consistently. The person's \n",
      "relative told us, \"There has been a lot of improvement in [person's] ulcer. I know this because she is no \n",
      "longer in pain.\"  A health professional told us, \"The wound has improved, there is strong evidence that the \n",
      "staff are turning [person] regularly and changing [person's] pads.\" Staff had also documented that the \n",
      "wounds 'showed signs of healing'. Staff we spoke to knew how to refer people to the district nurse and \n",
      "tissue viability nurse in order to obtain advice about the prevention and management of pressure ulcers \n",
      "when required. \n",
      "\n",
      "However, we found some inconsistencies in how identified risks were managed to ensure people stayed \n",
      "safe. We observed two occasions where people requiring supervision for certain tasks, were left unattended. \n",
      "One of these occasions was where a person required supervision whilst eating their meal. The person's care \n",
      "plan stated, \"Staff need to support [Person], as they have a tendency to put too much into their mouth.\" We \n",
      "spoke to a staff member nearby who told us that they knew the person required supervision but they were \n",
      "unable to do so fully as they had to attend to the needs of other people also. \n",
      "\n",
      "There were generic and environmental risk assessments for all areas of the home. Care had been taken to \n",
      "identify and manage risks around the home, but in some cases practice observed fell short of these control \n",
      "measures. We observed people walking into the open plan kitchen in flat one, with no staff supervision. The \n",
      "kitchen in flat three had a keypad entry system for the door, but this was observed to be left on the latch or \n",
      "wide open on several occasions during the inspection. The kitchens in all the flats contained small \n",
      "appliances such as microwaves, kettles and toasters. Records showed that the small appliances had been \n",
      "tested to ensure they were electrically safe to use, but there were no specific risk assessments to ensure that \n",
      "people who were able to enter the kitchens were safe to do so.\n",
      "\n",
      "We raised our concerns regarding mitigating risk management to the home management team who \n",
      "submitted an action plan shortly after the inspection to rectify the concerns we noted. \n",
      "\n",
      "At our last inspection, we found that medicines were not safely managed. At this inspection we found that \n",
      "the provider had addressed the issue. We looked at the systems in place for managing medicines. We spoke \n",
      "with seven staff involved in the governance and administration of medicines, observed medicine \n",
      "\n",
      "8 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "administration for five residents and examined 14 people's medicines administration records (MARs).\n",
      "\n",
      "Medicines were stored securely in medicine trollies which were kept in a locked room. We observed on \n",
      "occasions that staff administering medicines left the medicine trolley open and unattended in corridors. \n",
      "This was fed back to management who advised they would address the concerns with the identified staff. \n",
      "Staff recorded fridge and room temperatures daily which demonstrated that medicines were stored at the \n",
      "correct temperatures. The care home had a process for ordering and checking stock that made sure \n",
      "medicines were available for people. Medicines no longer required were clearly separated from other stock, \n",
      "recorded in a log and returned to the dispensing pharmacy.\n",
      "\n",
      "Medicines that require additional controls because of their potential for abuse (controlled drugs) were \n",
      "stored in a suitable cupboard. When a controlled drug was administered, the records showed the signature \n",
      "of the person administering the medicine and a witness signature. Staff completed regular stock checks. \n",
      "One controlled drug did not have the quantity recorded in the log book. This was highlighted to staff and the\n",
      "medicine was correctly recorded by the end of the inspection.\n",
      "\n",
      "Care staff administered medicines with care and patience and tailored the administration to the needs of \n",
      "the individual. We saw staff following a safe method for giving and recording medicines administration. The \n",
      "service provided regular medicine training and conducted competency assessments for staff.\n",
      "\n",
      "The dispensing pharmacy supplied the MARs and staff completed the MARs accurately to show what \n",
      "medicines people had taken. For medicines administered as and when required, there were clear protocols \n",
      "detailing how and when to give the medicine. Staff completed patch charts and topical administration \n",
      "records to show when medicines were applied to the skin. People's details and allergy status were recorded \n",
      "on the MARs and medicine profile. We saw one record where the medicine profile stated an allergy but the \n",
      "MAR stated no known allergies. This was highlighted to the management team on the day of inspection. \n",
      "They immediately addressed the concern and made sure that the person's allergy status was correct on all \n",
      "documentation.\n",
      "\n",
      "The procedures to enable people to self-administer medicines were not well established in the care home. \n",
      "We saw person that self-administered some of their medicines. While staff had completed a brief risk \n",
      "assessment, it did not comprehensively assess the person's ability to manage their medicines safely. The \n",
      "provider had recently introduced a medicine risk assessment as part of an overall documentation review. \n",
      "This was not yet in use at the home at the time of inspection. We raised this with the home management \n",
      "team who advised that they would ensure the risk assessment was completed by all responsible staff. \n",
      "\n",
      "The care home liaised regularly with the dispensing pharmacy and GP surgeries to discuss medicine \n",
      "management. The service strived for continuous improvement. The management had introduced a \n",
      "medicine audit programme which included daily and monthly audits. There was evidence that the audits \n",
      "were identifying areas for improvement and that action was taken to make the improvements.\n",
      "\n",
      "All people and relatives we spoke with told us they or their loved one were safe living at Apthorp Care \n",
      "Centre. Feedback received from people included, \"Got own room with a bed. They wash your clothes. You're \n",
      "safe here. Main thing in life is to be safe\", \"Yes they look after us well\" and \"Yes, so far I have not found \n",
      "anything to be afraid of.\" Relatives told us, \"[Person's] safer than when she was at home\", \"[Person] has been\n",
      "here three years and is happier now as it's much improved. They come and check on her more often\" and \"I \n",
      "think [person] feels safe, comfortable and loved.\" \n",
      "\n",
      "We checked to see if staff were safely recruited. Staff files included application forms, records of interview \n",
      "\n",
      "9 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "and appropriate references. Records showed that checks had been made with the Disclosure and Barring \n",
      "Service (criminal records check) to make sure that only suitable staff were employed to work at the home. \n",
      "Records seen confirmed that staff members were entitled to work in the UK.\n",
      "\n",
      "Staff had a good understanding of how to keep people safe and their responsibilities for reporting accidents,\n",
      "incidents or concerns. Records confirmed that staff had received training in safeguarding people. Staff also \n",
      "confirmed that they could access the safeguarding policy. They were able to describe the types of abuse to \n",
      "look out for and the steps they would take if they had concerns. Staff identified that they could report abuse \n",
      "concerns outside of the organisation to the local safeguarding authority and the Care Quality Commission \n",
      "(CQC). Staff were knowledgeable about what whistleblowing meant. A staff member told us, \"It's about \n",
      "keeping everyone safe. I would immediately report any concerns and if I was not happy with response I \n",
      "would go to head office and whistle blow. I could also go to the police and CQC. The management are pretty\n",
      "good with reporting.\"\n",
      "\n",
      "Accidents and incidents were recorded and actions and learning identified as a result of the incident were \n",
      "implemented. Staff knew how to report accidents and incidents. Falls were recorded and analysed on a \n",
      "monthly basis for trends and potential causes. Incidents such as falls and unexplained bruising or body \n",
      "marks were discussed by the home management team on a daily basis and monitored. Where appropriate, \n",
      "safeguarding alerts had been sent to the local safeguarding authority for further investigation. We saw that \n",
      "one person had been assessed at being at risk of climbing over their bed rails. As a result, the person's \n",
      "profiling bed had been lowered to its lowest point and had movement sensors installed. There were crash \n",
      "mattresses on both sides of their bed which allowed the person to slide on to the mattresses. Staff were \n",
      "knowledgeable around how to safely support the person during moving and handling. \n",
      "\n",
      "The safety of the building was routinely monitored and records showed appropriate checks and tests of \n",
      "equipment and systems such as fire alarms, emergency lighting and gas and electrical safety had been \n",
      "undertaken. The service also had contracts in place for the routine maintenance and servicing of \n",
      "equipment. Water temperatures were checked on a regular basis, however, throughout the inspection we \n",
      "observed that the hot water in some bedrooms and communal toilets was above the acceptable \n",
      "temperature range as stated by the Health and Safety Executive (HSE), which could have placed people at \n",
      "risk of scalding. We discussed our concerns with the home management team who commissioned an \n",
      "external service provider to review and recalibrate the water temperatures. The provider also assured that \n",
      "they would closely monitor the situation and undertake any necessary remedial works to ensure water \n",
      "temperatures were within an acceptable temperature range. \n",
      "\n",
      "Personal Emergency Evacuation Plans (PEEPs) were in place and the provider had a clear contingency plan \n",
      "in place to help ensure people were kept safe in the event of a fire or other emergency.\n",
      "\n",
      "People were protected by the use of safe infection control procedures and practices. Staff were trained and \n",
      "kept up to date with good practice. The home was overall clean and well maintained on the days we visited.\n",
      "\n",
      "10 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " Is the service effective?\n",
      "\n",
      "Requires Improvement\n",
      "\n",
      "Our findings  \n",
      "At our last inspection in February 2017, we found inconsistent recording on charts relating to people's \n",
      "health. People who were at high risk with regards to their nutrition and fluid intake had fluid and food \n",
      "monitoring charts to allow staff to monitor their food consumption and fluid intake. We found that some \n",
      "people's food and fluid charts were not regularly completed and contained gaps in recording. \n",
      "\n",
      "During this inspection, we found that the provider had not addressed the issue. Daily fluid intake targets \n",
      "were not identified for people at risk of dehydration. Fluid monitoring charts were inconsistently completed \n",
      "with gaps in recording seen, especially at night time. Although running fluid intake totals were kept, a daily \n",
      "total fluid intake was not calculated and where a person was recorded to have had a low fluid intake, it was \n",
      "not clear from review of documentation that any actions had been taken to escalate concerns regarding low\n",
      "fluid intake. We saw that the local authority had raised fluid intake monitoring with the provider on two \n",
      "occasions following the last inspection, however, the provider had failed to adequately implement a robust \n",
      "recording system for monitoring fluid intake for people who were at risk of dehydration. We observed that \n",
      "people had access to drinks and were regularly offered drinks by staff throughout the inspection which was \n",
      "reflected in feedback received from people. We observed that the head chef was knowledgeable of people \n",
      "at risk of dehydration and additional supplements to encourage hydration such as jellies, smoothies and \n",
      "fruits. Therefore the concerns noted are related to record-keeping as opposed to hydration. \n",
      "\n",
      "This was a continued breach of Regulation 17 of the Health and Social Care Act 2008 (Regulated Activities) \n",
      "Regulations 2014.\n",
      "\n",
      "We found that people's weights were proactively monitored using a Malnutrition Universal Screening Tool \n",
      "(MUST) which colour coded the level of risk of malnutrition posed to the person. Dependent on the level of \n",
      "risk, an action plan was implemented called 'food first strategy' where the person was supported with \n",
      "fortified foods and nutritional shakes. Where a person posed an increased risk, a referral to the dietician was \n",
      "made in addition to the 'food first strategy.' The registered manager maintained an overview system of \n",
      "people's monthly weights and we saw that some people at risk of malnutrition gained weight as a result of \n",
      "the interventions made. \n",
      "\n",
      "We received a mostly positive response when we asked people about the quality and choice of food on offer \n",
      "at Apthorp Care Centre. Comments received from people included, \"Food nice and you get a choice\", \"You \n",
      "get a choice. If you don't like you can have an omelette or jacket potato\", \"Sometimes they choose and \n",
      "sometimes they ask you what you want. The food is very good\" and \"A good average. But you don't expect A \n",
      "La Carte. They are pleasant at lunchtime.\" A relative told us, \"[Person] loves ice-cream and gets lots of it.\" \n",
      "\n",
      "We observed lunchtimes across all flats in the home on all three days of the inspection and found \n",
      "inconsistencies in the mealtime experience people received. In some flats where people were more able and\n",
      "could feed themselves, their lunchtime experience was calm and pleasant. In parts of the home where \n",
      "people required assistance with eating their meal, lunchtime was observed to be more chaotic as a result of \n",
      "poor staff deployment which was elaborated on in the safe section of this report. \n",
      "\n",
      "11 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "Although staff and some people told us people had a choice of meal, we did not always observe this in \n",
      "practice. We saw that there was one main meal on offer at lunch time. Alternatives such as baked potato \n",
      "and sandwiches were offered if someone did not want the main choice on offer. We saw that people made \n",
      "their lunch choice for the following day in the afternoon. However, we observed that when lunch was served,\n",
      "people were given the main meal only and alternatives were not offered if the person did not want the main \n",
      "choice. \n",
      "\n",
      "We observed that in most flats, the menu was on display either in the kitchen area away from people's view \n",
      "or not at all. When people asked what the meal they were being served was some staff could not tell them. \n",
      "For people living with dementia, they may not have been able to recall what they chose for their meal the \n",
      "day prior. We discussed our findings with the home management team who acknowledged that their meal \n",
      "time experience for people could be improved and they advised that they would make this a priority area for\n",
      "improvement. \n",
      "\n",
      "We observed some kind and patient interactions between staff and people who used the service at \n",
      "mealtimes. We saw that staff ate their meals alongside people at lunchtime and where people required \n",
      "prompting and assistance, staff provided this in a caring manner. We spoke to the chef who was very \n",
      "knowledgeable around people's dietary needs and preferences. \n",
      "\n",
      "At our previous inspections in July 2016 and February 2017, we found that the provider was not ensuring \n",
      "staff were adequately supported through regular supervision sessions. At this inspection, we found that the \n",
      "provider had made improvements in this area. \n",
      "\n",
      "People and their relatives spoke positively about staff and told us they were skilled to meet their needs. One \n",
      "person told us when asked, \"Definitely.\" A second person told us, \"Yes. Anybody I have asked has helped \n",
      "me.\" A third person told us, \"I'm happy to be with everyone\". A relative told us, \"Its obvious [staff] have had \n",
      "training.\" \n",
      "\n",
      "Staff told us they received regular supervision sessions with their line manager. A staff member told us, \"We \n",
      "discuss my role, what I am doing right/wrong, good/bad feedback which is very helpful.\" The provider was \n",
      "currently in the process of moving from a paper based supervision record to an electronic record which \n",
      "linked in with the appraisal process. The company policy stated that there should be two supervisions per \n",
      "year which is supplemented by a mid-year review and an annual appraisal. There was a large backlog of \n",
      "staff who had not received supervisions or appraisals in line with the frequency set out in the company \n",
      "policy, however, the service was working to reduce the backlog and expected to have everyone up to date by\n",
      "the end of the year. \n",
      "\n",
      "There was a cascade system for appraisals to ensure that they were spread out among senior staff and that \n",
      "the reliance would not be all on one senior staff member to undertake them with all staff. The registered \n",
      "manager supervised the deputy managers, who then supervised the floor managers who then supervised \n",
      "team leaders who then supervised support staff. \n",
      "\n",
      "The electronic system was set out so that there were set questions asked to all staff which linked in with \n",
      "their development and training needs and allowed staff to comment on their progress. Staff were asked to \n",
      "review the questions they would be asked in advance of the face to face meetings to allow them to consider \n",
      "their answers and reflect on their performance. New staff were given supervision more frequently when they \n",
      "joined the service to offer additional support and allowed them the opportunity to raise any learning needs.\n",
      "\n",
      "New employees attended a three day induction at head office prior to commencing work at the service. New\n",
      "\n",
      "12 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "staff then completed the care certificate over the first five weeks and spent a minimum of two weeks \n",
      "shadowing other staff. New staff were allocated to one particular flat to give them the opportunity to get to \n",
      "know the role and the people residing in that particular flat. Staff were not allowed to dispense medicines \n",
      "until they had passed their six month probationary period.\n",
      "\n",
      "All training was monitored centrally by the provider, however, the service also kept a spread sheet of staff \n",
      "training to ensure that the central register was correct and up to date. Staff received training in key areas \n",
      "relevant to their roles including moving and handling, infection control and health and safety. Staff were \n",
      "also offered training in addition to mandatory subjects such as diabetes care which was provided by \n",
      "external healthcare professionals. Staff were also supported to achieve additional vocational qualifications \n",
      "in health and social care which were provided by external providers. Staff told us they received regular \n",
      "training and were supported to obtain a national vocational qualification. \n",
      "\n",
      "The Mental Capacity Act 2005 (MCA) provides a legal framework for making particular decisions on behalf of \n",
      "people who may lack the mental capacity to do so for themselves. The Act requires that as far as possible \n",
      "people make their own decisions and are helped to do so when needed. When they lack mental capacity to \n",
      "take particular decisions, any made on their behalf must be in their best interests and as least restrictive as \n",
      "possible. People can only be deprived of their liberty to receive care and treatment when this is in their best \n",
      "interests and legally authorised under the MCA. The application procedures for this in care homes and \n",
      "hospitals are called the Deprivation of Liberty Safeguards (DoLS). Where people were deprived of their \n",
      "liberty, appropriate applications had been submitted to the local authority for DoLS assessments to be \n",
      "considered for authorisation. DoLS were monitored on an on-going basis and applications had been \n",
      "submitted for reviews and renewals in a timely manner. \n",
      "\n",
      "Staff had a good understanding of MCA and DoLS and understood the importance of asking for consent \n",
      "from people. Staff had attended MCA training. A staff member told us, \"A person without capacity needs \n",
      "safeguards in place to protect them such as DoLS. People can still make a choice by looking even if they \n",
      "can't understand.\" A second staff member told us if someone did not have mental capacity, \"I would still ask\n",
      "them what they would like to wear. I would still give them a choice.\" People told us that staff asked for their \n",
      "consent when providing assistance. A person told us, \"I have never found them [staff] intrusive.\" \n",
      "\n",
      "People were supported to maintain good health and had access to healthcare services and received on-\n",
      "going healthcare support. Care plans detailed records of appointments with health and care professionals. \n",
      "We also saw evidence that following appointments, people's care plans were updated accordingly. A person \n",
      "told us, \"Only recently I saw a Doctor.\" A second person told us, \"They take me in a wheelchair to see the \n",
      "Doctor.\" A relative told us, \"The GP has always been really accessible.\" We observed throughout the \n",
      "inspection that where staff had a concern about a person, they were prompt in arranging for the person to \n",
      "see a medical professional. In addition, we observed people's health appointments discussed at the \n",
      "morning handover meeting.  \n",
      "\n",
      "13 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " Is the service caring?\n",
      "\n",
      "Good\n",
      "\n",
      "Our findings  \n",
      "When we last inspected the service, we found that people were not always supported to maintain their \n",
      "dignity. At this inspection, we found that the provider had addressed the issue. Most people and all relatives \n",
      "told us they were happy with the care they received and spoke positively about staff. Comments received \n",
      "from people included, \"The staff are lovely\", \"Very happy here. Everybody is so nice and cheerful. Always \n",
      "here to help you\" and \"The staff are very caring and we are happy here. They really look after us. We are not \n",
      "short of anything.\" A relative told us, \"Staff are so conscientious. They really do care.\" A second relative told \n",
      "us, \"The atmosphere is very pleasant. Staff do as much for me as they do for [relative].\"\n",
      "\n",
      "We did receive some feedback from people stating that they felt that at times staff were rushed in \n",
      "completing their tasks and did not have much time to speak with people. A person told us, \"They are always \n",
      "working, always moving, no time to chat.\" However another person told us, \"Carers come and chat and talk \n",
      "when doing something.\" \n",
      "\n",
      "We observed kind and positive interactions between staff and the people they were supporting. Staff were \n",
      "gently orientating people when appropriate. Staff were maintaining some physical contact such as hand \n",
      "holding where this was appropriate and calming for the person.\n",
      "\n",
      "We spoke with people and relatives about whether they felt involved with their own care provision or the \n",
      "care provision of their relative. One relative told us, \"We talked through [person's] needs.\" \n",
      "\n",
      "Staff respected people's privacy and dignity. We saw that doors were kept closed when people were \n",
      "receiving personal care. People we spoke with told us they felt that they were treated with respect and that \n",
      "their privacy was protected. A person told us, \"I need help to get up and I get up when I want to. Staff ask \n",
      "permission.\" A relative told us, \"One day I came up and [person] was sick in the toilet. They [staff] were very \n",
      "tender with her. They always maintain her dignity throughout. Reassuring her the whole time.\" A staff \n",
      "member told us, \"Make sure the door is closed, encourage people to clean themselves and cover private \n",
      "areas whilst supporting with personal care.\"\n",
      "\n",
      "We saw that people were supported to maintain their independence, where appropriate. We saw one \n",
      "person make their own hot drinks and another person told us, \"I go to the church up the road.\" \n",
      "\n",
      "People's bedrooms were personalised with family photos, ornaments and pictures. A person told us, \"I got \n",
      "my own bed and chair and mum and dad's pictures up.\" We observed relatives visiting the home throughout\n",
      "the inspection with no restrictions on visiting times. A relative told us, \"No restriction. They are working hard \n",
      "to maintain a good friendly atmosphere.\" \n",
      "\n",
      "Care plans documented that advanced care planning and end of life care was discussed with most people \n",
      "and their relatives. People's choices and wishes were recorded in relation to planning the way in which they \n",
      "wanted to be cared. A relative told us, \"It is amazing and the staff are doing a wonderful job. You would not \n",
      "think that [person] is nearing the end of life because [person] is always very comfortable and well dressed. \n",
      "\n",
      "14 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "There is no foul smell although [person] is doubly incontinent. Most important of all [person] is not lonely, \n",
      "because the staff always walk in all the time.\" \n",
      "\n",
      "Staff we spoke to understood what equality and diversity meant and how that affected the care they \n",
      "provided for people who used the service. When asked how to work with people from a variety of \n",
      "backgrounds, staff told us, \"Care would not differ. It would be the same for everyone regardless.\" People's \n",
      "religious beliefs and cultural background were recorded in care files and people told us that they were able \n",
      "to attend religious services and local religious leaders visited them at the service on a regular basis.\n",
      "\n",
      "15 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " Is the service responsive?\n",
      "\n",
      "Good\n",
      "\n",
      "Our findings  \n",
      "People using this service and their relatives told us that the management and staff responded to any \n",
      "changes in their needs. We saw from people's care records and by talking with staff that if any changes to \n",
      "people's health were noted by staff, they would report these changes and concerns to the home \n",
      "management team and health professionals. Relatives told us they were kept up to date with any issues and\n",
      "involved in care reviews. A relative told us, \"[Person] knocked her hand and needed to be seen at hospital. \n",
      "They contacted us.\" A person who had recently been admitted with a pressure ulcer told us, \"My wound got \n",
      "seen very quickly by the nurse and treatment started in quick time. This is amazing.\" Visiting health \n",
      "professionals told us that staff were proactive in seeking advice and assistance when they had concerns. \n",
      "\n",
      "People's individual care plans included detailed information about their life history, dietary likes and \n",
      "dislikes, cultural and religious heritage, healthcare and medical needs, skincare and preferred routines. Care\n",
      "plans were reviewed regularly and updated as changes occurred.\n",
      "\n",
      "Through our observations and discussions with staff they demonstrated an awareness of people's \n",
      "preferences, what people were able to do and what they needed support with. Records showed \n",
      "preadmission information had been completed. An assessment was carried out to identify people's support \n",
      "needs and this included information about their medical conditions, behaviour, communication and their \n",
      "daily lives. \n",
      "\n",
      "People were supported to engage in activities on a daily basis. Comments from people regarding activities \n",
      "included, \"We went to Southend in July. It was a really lovely day\", \"Plenty to do. Quizzes, drawing, flowers. \n",
      "They take me out to have a walk\" and \"Yes they do and they also ask me to help out. Here today I am on the \n",
      "green balcony helping the activity coordinator prune the plants and tidy up.\" Some people we spoke with \n",
      "told us they did not want to take part in group activities and preferred to spend time in their rooms. During \n",
      "the inspection, we observed an activities co-ordinator visit people in their bedrooms on a one to one basis. \n",
      "\n",
      "Notices about activities and events were displayed around the home. We observed dedicated activities co-\n",
      "ordinators deliver the activities programme. On the days of the inspection, activities included poetry \n",
      "reading, gardening and bingo. In addition, on the first evening of the inspection, a cocktail party was held to \n",
      "celebrate the 25 year anniversary of the provider. We observed a game of bingo in the bar area on second \n",
      "morning.  About 18 people took part, mostly sitting round tables. Bingo was based on old time songs rather \n",
      "than numbers so much of the activity was based on singing which enabled those not able to play to nod and\n",
      "sing to the music. We noted that the activities coordinators knew people's names. \n",
      "\n",
      "We checked how the service handled complaints. We saw that any complaints received had been \n",
      "investigated fully and appropriately, in line with the home's policies and procedures. Responses from home \n",
      "management were open and detailed. Where mistakes had been made these were acknowledged and \n",
      "apologies provided. People and relatives generally told us that they had no complaints and any concerns \n",
      "raised were appropriately dealt with.\n",
      "\n",
      "16 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      " Is the service well-led?\n",
      "\n",
      "Requires Improvement\n",
      "\n",
      "Our findings  \n",
      "We received positive feedback from people and relatives regarding how the home was managed overall. \n",
      "Some people told us they did not know the registered manager, however, they knew who was in charge of \n",
      "their flat or floor and felt confident that they could raise any concerns or queries with them. Feedback \n",
      "received from people, relatives and staff was that the quality of care at Apthorp Care Centre had improved. \n",
      "Comments from people included, \"[registered manager] is easy to talk to, talked to him several times\" and \n",
      "\"If I had a complaint would leave it to my daughter. There is a lady here you can talk to. I know a woman \n",
      "manager but do not know a man\" and \"Yes it is a man.\"  \n",
      "\n",
      "A relative told us, \"Since new manager been in its got a lot better. [Person] has been here three years and is \n",
      "happier now as it's much improved. They come and check on her more often.\" A second relative told us, \n",
      "\"The rating went down and they deserved it. New manager came, [deputy care manager] came and things \n",
      "are better. It's how it should be run now. Easier to deal with issues now got manager on each floor. They ring\n",
      "me and keep me updated and you get continuity of staff. All good now.\" \n",
      "\n",
      "We received a mixed response from staff when asked about the management of the home. Staff were \n",
      "generally positive about the registered manager and the local management team; however, we found that \n",
      "some staff had low morale which was related to recent organisational changes with the provider and a lack \n",
      "of management support at busy times of the day. One staff member told us, \"We are not short staffed but \n",
      "around key times need support. We are not supported by managers. They are only around as CQC are \n",
      "around.\" A second staff member told us, \"Management is good.\" A professional involved with the service \n",
      "told us that they found that management support was not visible at mealtimes when they visited the \n",
      "service. We spoke to the home management team who told us that they would ensure increased managerial\n",
      "support at mealtimes and train non care staff to assist with some mealtime tasks. \n",
      "\n",
      "When we last inspected the service, we found that although quality monitoring measures were in place, they\n",
      "had not yet been embedded and did not pick up the concerns identified by the inspection team. At this \n",
      "inspection we found that improvements had been made to quality assurance was still not fully developed \n",
      "and embedded. In addition to a workbook, which was the required reporting tool submitted by the \n",
      "registered manager to the providers head office every month, we saw that the registered manager had \n",
      "created a monthly audit unique to Apthorp Care Centre which was completed by the floor manager on each \n",
      "floor. The audit identified areas for improvement such as health and safety, maintenance, DoLS monitoring \n",
      "and care plan reviews. We saw that the registered manager reviewed the monthly audits and made \n",
      "comments regarding improvements to be actioned. The registered manager told us that the audit findings \n",
      "were discussed with the floor manager in a supervision session as opposed to on a monthly basis to ensure \n",
      "improvements needed were actioned promptly. \n",
      "\n",
      "When we last inspected the service, we found concerns relating to fluid monitoring. At this inspection we \n",
      "found sufficient improvement had not been made. \n",
      "\n",
      "We saw that a quality assurance survey was completed in July 2017. 50 questionnaires had been completed \n",
      "\n",
      "17 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "and in eight of the questionnaires, negative feedback had been received. We saw no evidence of an action \n",
      "plan or how concerns or negative comments would be addressed. The registered manager stated that all \n",
      "responses were sent to head office for analysis and an action plan  was expected to be compiled in October \n",
      "2017. However the regional director advised us that negative feedback would be addressed immediately. \n",
      "Overall, we found that although there were systems and processed in place for effective quality monitoring, \n",
      "these had not yet been embedded. \n",
      "\n",
      "The service had a home improvement plan in place. Findings from previous CQC inspections and local \n",
      "authority feedback formed the basis of the home improvement plan. The home improvement plan was \n",
      "updated when actions had been completed and examples of recently completed improvement actions, \n",
      "which were evidenced on inspection, included medicines management and auditing of accidents, incidents \n",
      "and falls. \n",
      "\n",
      "Staff meetings took place on a regular basis. Topics discussed at meetings included CQC compliance, \n",
      "medicines and importance of consent. In addition to staff meetings, a daily handover and a '10 at 10' \n",
      "meeting took place. At the handover, feedback from night staff was relayed to the care manager who \n",
      "advised day staff of any concerns or issues. We were advised that the handover process would be changing \n",
      "and a detailed handover document would shortly be trialled in two flats. This handover document would  be\n",
      "individual to each person and would include their background history, observations and care \n",
      "recommendations. \n",
      "\n",
      "Residents and relatives meetings took place on a regular basis. Topics discussed included communication \n",
      "within the home, CQC inspection findings and actions taken to ensure improvement, role of local \n",
      "authorities, social events and activities. \n",
      "\n",
      "Throughout the inspection we gave feedback to the home management team and senior management and \n",
      "clarification was sought where necessary, for example in relation to improvements to medicines \n",
      "management and staffing concerns at mealtimes. The registered manager and deputy manager \n",
      "demonstrated a willingness to learn and reflect in order to improve the service people received as a result. \n",
      "Shortly after the inspection, the registered manager submitted an action plan to the lead inspector outlining\n",
      "actions to be taken to address the issues identified throughout this inspection.\n",
      "\n",
      "18 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "This section is primarily information for the provider\n",
      "\n",
      "Action we have told the provider to take\n",
      "\n",
      "The table below shows where regulations were not being met and we have asked the provider to send us a \n",
      "report that says what action they are going to take.We will check that this action is taken by the provider.\n",
      "\n",
      "Regulated activity\n",
      "Accommodation for persons who require nursing or \n",
      "personal care\n",
      "\n",
      "Regulation\n",
      "Regulation 17 HSCA RA Regulations 2014 Good \n",
      "governance\n",
      "\n",
      "Regulated activity\n",
      "Accommodation for persons who require nursing or \n",
      "personal care\n",
      "\n",
      "Regulation 17(1) \n",
      "\n",
      "The provider had not ensured that health \n",
      "monitoring forms had been completed for \n",
      "people at risk of dehydration.\n",
      "\n",
      "Regulation\n",
      "Regulation 18 HSCA RA Regulations 2014 Staffing\n",
      "\n",
      "Regulation 18(1) \n",
      "\n",
      "The provider had not ensured sufficient staff \n",
      "were effectively deployed at mealtimes.\n",
      "\n",
      "19 Apthorp Care Centre Inspection report 04 January 2018\n",
      "\n",
      "\f",
      "\n"
     ]
    }
   ],
   "source": [
    "from pdfminer.pdfinterp import PDFResourceManager, PDFPageInterpreter\n",
    "from pdfminer.converter import TextConverter\n",
    "from pdfminer.layout import LAParams\n",
    "from pdfminer.pdfpage import PDFPage\n",
    "from io import StringIO\n",
    "\n",
    "# create a function to read a pdf document and conver to text, returns text\n",
    "\n",
    "def convert(fname, pages=None):\n",
    "# cant get pages func to run...\n",
    "\n",
    "    if not pages:\n",
    "        pagenums = set()\n",
    "    else:\n",
    "        pagenums = set(pages)\n",
    "\n",
    "    output = StringIO()\n",
    "    manager = PDFResourceManager()\n",
    "    converter = TextConverter(manager, output, laparams=LAParams())\n",
    "    interpreter = PDFPageInterpreter(manager, converter)\n",
    "\n",
    "    infile = open(fname, 'rb')\n",
    "    \n",
    "    for page in PDFPage.get_pages(infile, pagenums):\n",
    "        interpreter.process_page(page)\n",
    "    infile.close()\n",
    "    converter.close()\n",
    "    text = output.getvalue()\n",
    "    output.close\n",
    "    \n",
    "    return text \n",
    "\n",
    "# create an object using the converted text, use the file path for the data\n",
    "text_file = convert('Data/sample.pdf')\n",
    "\n",
    "print(text_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### read and convert multiple pdf files, create a dataframe of text files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                   0\n",
      "0  miss s g howard victoria lodge care home date ...\n",
      "1  healthcare homes lsc limited sandown park care...\n",
      "2  the cedars baildon limited the cedars inspecti...\n",
      "3  valeo limited the lodge inspection report scar...\n",
      "4  summerfield private residential home limited s...\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import glob\n",
    "from sklearn.model_selection import train_test_split\n",
    "import re\n",
    "\n",
    "files = glob.glob(\"Data/*reports/*.pdf\")\n",
    "\n",
    "records = []\n",
    "\n",
    "for i in files:\n",
    "\n",
    "    #open the file\n",
    "    text_file = convert(i)\n",
    "\n",
    "    # read and process text\n",
    "    text = text_file.replace(\"\\n\", \" \")\n",
    "\n",
    "#     # lets remove the classes from the text in case they are infliuencing the tests\n",
    "#     to_remove = ['good', 'outstanding', 'inadequate', 'improve']\n",
    "\n",
    "#     for term in to_remove:\n",
    "#         text = text.replace(term, ' ')\n",
    "\n",
    "    text = re.sub('[^a-zA-Z]+', ' ', text)\n",
    "    # make it all lower case\n",
    "    text = text.lower()\n",
    "\n",
    "    # append to our list of reports\n",
    "    records.append(text)\n",
    "\n",
    "#convert to dataframe\n",
    "df = pd.DataFrame(records)\n",
    "\n",
    "#write train and test data\n",
    "df.to_csv(\"Data/full_text.txt\", header = None, index= False, quotechar=\" \")\n",
    "\n",
    "# print the head to show a couple of rows\n",
    "print(df.head())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### read in sections from from the pdf reports "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Overall rating for this service  \\\n",
      "0                     outstanding   \n",
      "1                     outstanding   \n",
      "2            requires_improvement   \n",
      "3            requires_improvement   \n",
      "4                            good   \n",
      "\n",
      "                                     Overall summary  \\\n",
      "0  Miss S G Howard Victoria Lodge Care Home  Date...   \n",
      "1  Healthcare Homes (LSC) Limited Sandown Park Ca...   \n",
      "2  The Cedars (Baildon) Limited The Cedars  Inspe...   \n",
      "3  Valeo Limited The Lodge  Inspection report  20...   \n",
      "4  Summerfield Private Residential Home Limited S...   \n",
      "\n",
      "                                Is the service safe?  \\\n",
      "0    Is the service safe?  Is the service effecti...   \n",
      "1    Is the service safe?  Is the service effecti...   \n",
      "2    Requires Improvement    Is the service safe?...   \n",
      "3    Requires Improvement    Is the service safe?...   \n",
      "4    Is the service safe?  Is the service effecti...   \n",
      "\n",
      "                           Is the service effective?  \\\n",
      "0    This unannounced inspection took place on th...   \n",
      "1    Our inspection took place on 9 January and 1...   \n",
      "2    The Cedars is a 'care home' in Baildon, Brad...   \n",
      "3    This inspection of The Lodge took place on 1...   \n",
      "4    Summerfield Private Residential Home is a ca...   \n",
      "\n",
      "                              Is the service caring?  \\\n",
      "0    The service was safe  People told us they fe...   \n",
      "1    The service was safe.  Effective systems wer...   \n",
      "2    The service was not consistently safe.   Peo...   \n",
      "3    The service was safe.  Risk assessments were...   \n",
      "4    The service was safe.  People told us they f...   \n",
      "\n",
      "                          Is the service responsive?  \\\n",
      "0    The service was effective.   There were arra...   \n",
      "1    The service was effective.  People's likes, ...   \n",
      "2    The service was not consistently effective. ...   \n",
      "3    The service was not always effective.  Staff...   \n",
      "4    The service was effective.  People received ...   \n",
      "\n",
      "                            Is the service well-led?  \\\n",
      "0    The service was very caring.  People receive...   \n",
      "1    Outstanding    4 Sandown Park Care Home Insp...   \n",
      "2    The service was caring.   4 The Cedars Inspe...   \n",
      "3    The service was caring.   Staff interacted w...   \n",
      "4    The service was caring.  People and relative...   \n",
      "\n",
      "                       Background to this inspection  \\\n",
      "0    The service was extremely responsive.    Peo...   \n",
      "1                 The service remains outstanding.     \n",
      "2    The service was responsive.  People received...   \n",
      "3    The service was responsive.  People's care p...   \n",
      "4    The service was responsive.  People spoke po...   \n",
      "\n",
      "                                Is the service safe?  \\\n",
      "0    The service was very well led.  People lived...   \n",
      "1    The service was well-led.  People and relati...   \n",
      "2    The service was not consistently well led.  ...   \n",
      "3    The service was not always well-led.  The se...   \n",
      "4    The service was well led.  People and relati...   \n",
      "\n",
      "                                        Our findings  \\\n",
      "0   We carried out this inspection under Section ...   \n",
      "1   We carried out this inspection under Section ...   \n",
      "2   We carried out this inspection under Section ...   \n",
      "3   We carried out this inspection under Section ...   \n",
      "4   We carried out this inspection under Section ...   \n",
      "\n",
      "                         ...                          \\\n",
      "0                        ...                           \n",
      "1                        ...                           \n",
      "2                        ...                           \n",
      "3                        ...                           \n",
      "4                        ...                           \n",
      "\n",
      "                                        Our findings  \\\n",
      "0     People and relatives all gave us very posit...   \n",
      "1     People and relatives provided positive feed...   \n",
      "2     People were kept safe from abuse and improp...   \n",
      "3     People told us they felt safe at The lodge....   \n",
      "4     People told us they felt safe living at Sum...   \n",
      "\n",
      "     Is the service caring?  \\\n",
      "0                    Good     \n",
      "1                    Good     \n",
      "2    Requires Improvement     \n",
      "3    Requires Improvement     \n",
      "4                    Good     \n",
      "\n",
      "                                        Our findings  \\\n",
      "0     People and relatives we spoke with told us ...   \n",
      "1     People and relatives told us they felt staf...   \n",
      "2     We found people's care needs were met by th...   \n",
      "3     Relatives told us they felt confident staff...   \n",
      "4     Care records demonstrated people's needs we...   \n",
      "\n",
      "  Is the service responsive?  \\\n",
      "0              Outstanding     \n",
      "1              Outstanding     \n",
      "2                     Good     \n",
      "3                     Good     \n",
      "4                     Good     \n",
      "\n",
      "                                        Our findings Is the service well-led?  \\\n",
      "0     People living in Victoria Lodge were happy ...                   Good     \n",
      "1     The service continued to provide outstandin...            Outstanding     \n",
      "2     People's feedback about staff was overwhelm...                   Good     \n",
      "3     People told us the staff were caring. One p...                   Good     \n",
      "4     At the last two inspections we found a lack...                   Good     \n",
      "\n",
      "                                        Our findings  \\\n",
      "0     Relatives told us that the supportive and r...   \n",
      "1     The service continued to provide outstandin...   \n",
      "2     People and relatives all praised the standa...   \n",
      "3     People told us they were supported to make ...   \n",
      "4     People told us there was a personalised app...   \n",
      "\n",
      "  Action we have told the provider to take  \\\n",
      "0                            Outstanding     \n",
      "1                            Outstanding     \n",
      "2                   Requires Improvement     \n",
      "3                   Requires Improvement     \n",
      "4                                   Good     \n",
      "\n",
      "                                                   j  \\\n",
      "0     The registered manager and provider promote...   \n",
      "1     The service's management and care team were...   \n",
      "2     Although we recognised the home was small a...   \n",
      "3     People told us the service was well-led. On...   \n",
      "4     People and relatives told us they were high...   \n",
      "\n",
      "                                                   k  \n",
      "0     The registered manager and provider promote...  \n",
      "1     The service's management and care team were...  \n",
      "2    The table below shows where regulations were...  \n",
      "3    The table below shows where regulations were...  \n",
      "4     People and relatives told us they were high...  \n",
      "\n",
      "[5 rows x 21 columns]\n"
     ]
    }
   ],
   "source": [
    "import nltk\n",
    "import re\n",
    "import pprint\n",
    "from nltk import word_tokenize\n",
    "import pandas as pd\n",
    "import glob\n",
    "\n",
    "files = glob.glob(\"Data/*reports/*.pdf\")\n",
    "\n",
    "# classes = ['outstanding','good','requires_improvement','inadequate']\n",
    "\n",
    "sections = ['Overall rating for this service',\n",
    "            'Overall summary',\n",
    "            'Is the service safe?',\n",
    "            'Is the service effective?',\n",
    "            'Is the service caring?',\n",
    "            'Is the service responsive?',\n",
    "            'Is the service well-led?',\n",
    "            'Background to this inspection',\n",
    "            'Is the service safe?',\n",
    "            'Our findings',\n",
    "            'Is the service effective?',\n",
    "            'Our findings',\n",
    "            'Is the service caring?',\n",
    "            'Our findings',\n",
    "            'Is the service responsive?',\n",
    "            'Our findings',\n",
    "            'Is the service well-led?',\n",
    "            'Our findings',\n",
    "            'Action we have told the provider to take']\n",
    "\n",
    "records = []\n",
    "\n",
    "# for all the files\n",
    "for i in files:\n",
    "\n",
    "    #split the file path/name to get the class names as defined in the class list above\n",
    "    a,b = i.split(\"Data/\")\n",
    "    c,d = b.split(\"_reports\")\n",
    "\n",
    "    #open the file\n",
    "    text_file = convert(i)\n",
    "\n",
    "    # read and process text\n",
    "    text = text_file.replace ('\\n', ' ')\n",
    "\n",
    "    #empty list to contain the sections of each file\n",
    "    split_text = []\n",
    "    counter = 0\n",
    "    split_text.append(c)\n",
    "    # split text by each of the items in the section list defined above\n",
    "    for j in range(0,len(sections)):\n",
    "        split_text.append(text.split(sections[j],1)[0])\n",
    "        counter =+ 1\n",
    "        text = text.split(sections[j],1)[-1]\n",
    "\n",
    "    #append the split string to the split list\n",
    "    split_text.append(text.split(sections[j],1)[-1])\n",
    "\n",
    "    #append the split list to records (where a record contains a list of section texts for a report. )\n",
    "    records.append(split_text)\n",
    "\n",
    "#print(records[1][6])\n",
    "#turn this list of lists into a dataframe /, columns=list(sections)\n",
    "cols = sections + ['j','k']\n",
    "df = pd.DataFrame(records, columns=list(cols))\n",
    "print(df.head())\n",
    "\n",
    "# print out to tsv (tab seperated) for use elsewhere\n",
    "df.to_csv(\"Data/structured_data.csv\", header=True, sep=\"\\t\")\n",
    "dfhead = df.head(20)\n",
    "dfhead.to_csv(\"Data/structured_data_head.csv\", header=True, sep=\"\\t\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## NLP Basics\n",
    "\n",
    "now that we have our pdf files converted into text and saved as a dataframe, we can use NLP techniques to prepare the data\n",
    "\n",
    "here's some defintions of words we'll be seeing a lot of in NLP:\n",
    "\n",
    "> *Tokenization* – process of converting a text into tokens\n",
    "\n",
    "> *Tokens* – words or entities present in the text\n",
    "\n",
    "> *Text object* – a sentence or a phrase or a word or an article\n",
    "\n",
    "lets go through some basic NLP functions that will help us....\n",
    "\n",
    "\n",
    "| Table of Contents |\n",
    "| ----------------- |\n",
    "| Stemming          |\n",
    "| Lemmatisation |\n",
    "| Word Embeddings |\n",
    "| Part-of-Speech Tagging |\n",
    "| Named Entity Disambiguation |\n",
    "| Named Entity Recognition |\n",
    "| Sentiment Analysis |\n",
    "| Semantic Text Similarity |\n",
    "| Language Identification |\n",
    "| Text Summarisation |"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nltk \n",
    "\n",
    "# nltk.download()\n",
    "\n",
    "### prompts for user to select corpora to install - indicate which are required"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Noise Removal\n",
    "\n",
    "Any piece of text which is not relevant to the context of the data and the end-output can be specified as the noise.\n",
    "\n",
    "For example:\n",
    "- language stopwords (commonly used words of a language – is, am, the, of, in etc), \n",
    "- URLs or links, social media entities (mentions, hashtags), \n",
    "- punctuations and industry specific words. \n",
    "\n",
    "This step deals with removal of all types of noisy entities present in the text.\n",
    "\n",
    "A general approach for noise removal is to prepare a dictionary of noisy entities, and iterate the text object by tokens (or by words), eliminating those tokens which are present in the noise dictionary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'at the sample text'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "noise_list = [\"is\", \"a\", \"this\", \"...\"] \n",
    "\n",
    "def _remove_noise(input_text):\n",
    "    words = input_text.split() \n",
    "    noise_free_words = [word for word in words if word not in noise_list] \n",
    "    noise_free_text = \" \".join(noise_free_words) \n",
    "    return noise_free_text\n",
    "\n",
    "_remove_noise(\"this is a at the ... sample text\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### StopWords are words which do not contain important significance to be used in text mining."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus import stopwords\n",
    "\n",
    "#Usually these stopwords are filtered out because they return vast amount of unnecessary information\n",
    "stopWords = set(stopwords.words('english'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### sometimes we dont need punctuation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'It is easy to get our hands on millions  of words of text  What can we do with it '"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import string\n",
    "\n",
    "#set of punctuation characters\n",
    "pn = set(string.punctuation + '—')\n",
    "\n",
    "#function to remove punctuation characters\n",
    "\n",
    "def unpunkt(input):\n",
    "    output = ''.join(ch for ch in input if ch not in pn)\n",
    "    return output\n",
    "\n",
    "\n",
    "#lets test it.\n",
    "\n",
    "\n",
    "punkttext= '''It is easy to get our hands on millions !!! of words of #text . What can we do with it? '''\n",
    "unpunkt(punkttext)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'F M S Scunthorpe Limited\\nAmber House\\n\\nInspection report\\n\\n12 Scawby Road\\nBroughton\\nBrigg\\nLincolnshire\\nDN20 0AF\\n\\nTel 01652657349\\n\\nRatings\\n\\nDate of inspection visit\\n17 August 2017\\n18 August 2017\\n\\nDate of publication\\n14 November 2017\\n\\nOverall rating for this service\\n\\nInadequate  \\n\\nIs the service safe\\n\\nIs the service effective\\n\\nIs the service caring\\n\\nIs the service responsive\\n\\nIs the service wellled\\n\\nInadequate     \\n\\nRequires Improvement     \\n\\nRequires Improvement     \\n\\nRequires Improvement     \\n\\nInadequate     \\n\\n1 Amber House Inspection report 14 November 2017\\n\\n\\x0cSummary of findings\\n\\nOverall summary\\n\\nAmber House is registered with the Care Quality Commission CQC to provide accommodation and \\npersonal care for a maximum of 13 people It is situated in the village of Broughton close to local amenities\\n\\nAccommodation is provided over two floors in single bedrooms with en suite facilities There is a large \\nsitting room set out into two separate areas a small sitting area and a diningactivities area Two further \\nlounges are situated on the ground floor Enclosed gardens to the side and rear of the building are easily \\naccessible There is also a supported living house called Redbourne House We undertook this inspection on\\n17 and 18 August 2017 The last inspection was carried out on 29 and 30 March and 3 April 2017 where the \\nservice was found to be noncompliant with the regulations 8 9 10 11 12 13 15 16 17 18 and 19 of the \\nHealth and Social Care Act 2008 Regulated Activities as well as regulation 18 of the registration regulations\\nDue to concerns found during the inspection the overall rating for the service was Inadequate and the \\nservice was therefore in special measures \\n\\nServices in special measures will be kept under review and if we have not taken immediate action to \\npropose to cancel the providers registration of the service will be inspected again within six months The \\nexpectation is that providers found to have been providing inadequate care should have made significant \\nimprovements within this timeframe\\n\\nIf not enough improvement is made within this timeframe so that there is still a rating of inadequate for any \\nkey question or overall we will take action in line with our enforcement procedures to begin the process of \\npreventing the provider from operating this service This will lead to cancelling their registration or to varying\\nthe terms of their registration within six months if they do not improve This service will continue to be kept \\nunder review and if needed could be escalated to urgent enforcement action Where necessary another \\ninspection will be conducted within a further six months and if there is not enough improvement so there is \\nstill a rating of inadequate for any key question or overall we will take action to prevent the provider from \\noperating this service This will lead to cancelling their registration or to varying the terms of their \\nregistration\\n\\nFor adult social care services the maximum time for being in special measures will usually be no more than \\n12 months If the service has demonstrated improvements when we inspect it and it is no longer rated as \\ninadequate for any of the five key questions it will no longer be in special measures\\n\\nFollowing the inspection in March and April 2017 we met with the provider and have requested an interim \\naction plan We also requested and have received weekly updates to assure us actions have been taken to \\naddress the concerns \\n\\nAfter the focused inspection March and April 2017 the provider contracted the support of a management \\ncompany to help them make the required improvements and achieve compliance with the regulations We \\nmet with the consultant who represented the management company at this inspection We refer to them as\\n\\n2 Amber House Inspection report 14 November 2017\\n\\n\\x0cthe consultant throughout this report\\n\\nDuring this inspection we found the providers quality assurance systems continued to be ineffective The \\ngovernance systems operated by the provider failed to cover all aspects of care delivery and did not identify \\nareas of the service that required improvement The provider did not have oversight of the service as \\nrequired and was not consulted regarding peoples care and support\\n\\nPeople who used the service did not always receive their medicines as prescribed Instructions to staff about\\nwhen medicines should be administered were not clear \\n\\nStaff had not always completed the necessary training to deliver the care and support the people who used \\nthe service required\\n\\nPeoples care plans did not always contain suitable guidance to ensure staff could meet their needs \\neffectively and consistently\\n\\nThe providers recruitment practices were not robust Some staff files did not contain relevant recruitment \\ninformation and thorough checks were not completed before agency staff worked with the people who used\\nthe service\\n\\nElements of the environment did not meet peoples needs we found a window that did not have a restrictor\\nradiator covers were not in place and wooden garden furniture was water damaged with split and pliable \\nwood\\n\\nWe found that people had restrictions in their lives that had not been agreed as in their best interests \\n\\nThe registered manager did not have access to some peoples care records and failed to ensure they have \\noversight of their care This meant there could be delays in any care or treatment they required\\n\\nWe saw that improvements had been made with regards to some of the risks we identified at our previous \\ninspection but found further evidence that the provider was not mitigating risks as required For example \\nlegionella checks were not being carried out which put people who used the service at unnecessary risk We \\nalso found items of food which indicated on the packet that they could not be frozen in the freezer and \\nitems that were past their used by or best before dates in fridges\\n\\nStaff who received adequate support and mentorship were deployed in suitable numbers to meet peoples \\nneeds Staff had completed safeguarding training and were aware of their responsibilities to report any \\npotential abuse and episodes of poor care they became aware of\\n\\nPeople were supported to eat a diet of their choosing and encouraged to make healthy choices\\n\\nPeople were treated with dignity and respect by staff they were enabled to make choices in their daily lives \\nWe saw people taking part in different activities throughout the inspection\\n\\nThe service had a registered manager who fulfilled their regulatory obligations to inform the Commission of \\nnotifiable events that occurred with the service\\n\\n3 Amber House Inspection report 14 November 2017\\n\\n\\x0cInadequate  \\n\\nThe five questions we ask about services and what we found\\n\\nWe always ask the following five questions of services\\n\\nIs the service safe\\n\\nThe service was not always safe People did not always receive \\ntheir medicines as prescribed \\n\\nRecruitment practices were not robust and the suitability of \\nagency staff was not adequately assessed\\n\\nRisks were not appropriately mitigated to ensure the health \\nsafety and welfare of the people who used the service\\n\\nStaff were deployed in suitable numbers to meet the needs of the\\npeople who used the service\\n\\nStaff had been trained to recognise the signs that could indicate \\nabuse had occurred and understood their responsibilities to \\nreport any poor care they became aware of\\n\\nIs the service effective\\n\\nRequires Improvement  \\n\\nThe service was not always effective People who used the \\nservice had restrictions placed upon them that had not been \\nagreed as in their best interests \\n\\nPeople were not always supported by staff who had the skills and\\nabilities to meet their needs effectively\\n\\nSome parts of the service had not been adapted effectively to \\nmeet people needs\\n\\nStaff received regular supervision and annual appraisals\\n\\nIs the service caring\\n\\nRequires Improvement  \\n\\nThe service was not always caring Some peoples care records \\nwere not stored appropriately and the registered manager did \\nnot have access to them \\n\\nPeoples needs were met by attentive and caring staff \\n\\nPeople were treated with dignity and respect by staff However \\nwe saw staff had used inappropriate language when describing \\nsome care interventions\\n\\n4 Amber House Inspection report 14 November 2017\\n\\n\\x0cPeople were supported to make choices in their daily lives\\n\\nIs the service responsive\\n\\nRequires Improvement  \\n\\nThe service was not always responsive Although some peoples \\ncare plans had been updated since our last inspection we found \\nthat some contained generic information and lacked appropriate\\nguidance for staff to follow\\n\\nThe registered provider had a complaints policy and procedure \\nin place The procedure was available in an easy read format to \\nmake it more accessible to the people who used the service \\n\\nPeople who used the service were encouraged to take part in \\nactivities of their choosing\\n\\nIs the service wellled\\n\\nInadequate  \\n\\nThe service was not always wellled The governance systems \\nand processes at the service were not effective\\n\\nThe provider did not take responsibility for the day to day \\nmanagement of the service \\n\\nThere was a registered manager in place and the CQC received \\nnotifications as required\\n\\n5 Amber House Inspection report 14 November 2017\\n\\n\\x0c \\n\\nAmber House\\n\\nDetailed findings\\n\\nBackground to this inspection\\nWe carried out this inspection under Section 60 of the Health and Social Care Act 2008 as part of our \\nregulatory functions This inspection was planned to check whether the provider is meeting the legal \\nrequirements and regulations associated with the Health and Social Care Act 2008 to look at the overall \\nquality of the service and to provide a rating for the service under the Care Act 2014\\n\\nThe inspection took place on the 17 and 18 August 2017 and was unannounced The inspection was \\ncompleted by three adult social care inspectors\\n\\nBefore this inspection we reviewed the information we held about the service such as notifications we had \\nreceived from the provider and their actions plans submitted to us following the last inspection \\nNotifications are when providers send us information about certain changes events or incidents that occur \\nWe spoke with the local authority commissioners and the local authority safeguarding team\\n\\nDuring the inspection we spoke with the registered manager five members of care staff and two members of\\ndomestic staff We also spoke with four people who used the service and three visiting healthcare \\nprofessionals The consultant from the management company the provider had commissioned to support \\nwith the required improvements at the service was also spoken with\\n\\nWe looked at six peoples care plans along with the associated risk assessments and Medication \\nAdministration Records MARs We also looked at how the service used the Mental Capacity Act 2005 and \\nDeprivation of Liberty Safeguards DoLS to ensure that when people were assessed as lacking capacity to \\nmake informed decisions themselves or when they were deprived of their liberty actions were taken in their \\nbest interest\\n\\nWe also looked at a selection of documentation pertaining to the management and running of the service \\nThis included action plans policies and procedures business continuity plans personal emergency \\nevacuation plans recruitment information for staff staff training records as well as audits and quality \\nmonitoring information\\n\\n6 Amber House Inspection report 14 November 2017\\n\\n\\x0c Is the service safe\\n\\nInadequate\\n\\nOur findings  \\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 12 2 g of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 safe care \\nand treatment We found peoples prescribed medicines including PRN as required medicines were not \\nrecorded on their medicines administration records MARs We were unable to find information when the \\nmedicines should be administered or when they had been given We also found protocols did contain clear \\ninstructions and suitable guidance to ensure they were administered safely and consistently\\n\\nDuring this inspection we saw one person had not received a medicine as prescribed They had been \\nprescribed a medicine in patch form to be applied every 72 hours However staff had consistently applied \\nthe patch every 96 hours This had not been identified by any of the staff team during the timeframe of \\nrecords we checked which was between 10 July and 17 August 2017 We have asked the registered manager\\nto discuss this with the persons GP\\n\\nStaff had recorded they had given PRN medicines for pain but not what the effect had been Staff had \\nhandwritten some MARs however full instructions about medicines had not been transcribed and a counter\\nsignature to help prevent mistakes had not always been made Some medicines had been written by the \\nprescriber as as directed this did not give clear instruction to staff and we have asked the registered \\nmanager to discuss this with the prescriber Unclear directions posed a risk that errors may occur and \\npeople may not receive the medicines they had been prescribed\\n\\nWe saw protocols were in place for staff guidance when people who used the service were prescribed \\nmedicines to be administered when required referred to as PRN The protocols did not give full \\ninformation to staff For example some people were prescribed a medicine to reduce anxieties up to two or \\nfour times a day when required however there was no information about the interval required between the \\ndoses Other protocols referred to staff administering medicines when required when people displayed \\nsigns of constipation displayed signs of suffering or for outward signs of agitation and anxiety there was \\nno information in the protocols about what these signs were for individual people We also saw that some \\npeople were prescribed medicines as a variable dose for example one to two to be taken at night There was\\nno information to guide staff regarding the decisionmaking about which dose to administer \\n\\nThis demonstrated a continuing breach of Regulation 12 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 safe care and treatment The action we have asked the provider to take can be \\nfound at the end of this report\\n\\nPeople who used the service told us they received their medicines on time One person said I always get it \\non time I have them at night too\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 19 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 We looked at \\nthree staff recruitment files and found two had positive disclosure and barring service DBS checks One of \\n\\n7 Amber House Inspection report 14 November 2017\\n\\n\\x0cthese staff members had not declared the offences on their application form contrary to the providers \\nrecruitment policy which should preclude them from working at the service\\n\\nDuring this inspection we were told by the consultant that there had been a number of staff had left the \\nprovider recently They said We have made a lot of change since the last inspection and some staff have \\nleft which we view as a positive step We have kept all of the ones we wanted to keep and lost the ones we \\nwere happy to see go The registered manager added The staff we have now are dedicated and very \\ncompassionate \\n\\nWe asked how the provider ensured they deployed sufficient staff and were told that agency staff were being\\nused as an interim measure The consultant explained We are currently using agency staff The agency \\nprovides all of the relevant information regarding staff that we need\\nWe reviewed the information held by the service in relation to agency staff and found that they could not \\nprovide evidence that suitable checks had been undertaken There was no evidence to support that \\nreferences had been gained by the agency or that gaps in peoples employment history had been explored \\n\\nWe assessed the recruitment of three permanent staff and found that interview questions and responses \\nwere not always available \\n\\nThis demonstrated a continuing breach of Regulation 19 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 fit and proper persons employed The action we have asked the provider to take\\ncan be found at the end of this report\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 12 2 a b of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 safe \\ncare and treatment We found the provider failed to appropriately assess and mitigate known risks in areas \\nincluding but not limited to buildings works being carried out at the service emergency evacuation plans \\nas well as fire and building safety \\n\\nDuring this inspection we found improvements had been made to the risks we had previously identified but \\nidentified other risks that had not been appropriately mitigated \\n\\nBuilding works had been completed and no longer posed at risk to the people who used the service We saw\\nevidence that the local fire safety officer had visited and due to improvements made following our \\ninspection had no concerns regarding fire safety management at the service We saw personal emergency \\nevacuation plans PEEPs had been updated for some people but not all This meant they were at risk of not \\nreceiving the support they required in an emergency situation \\n\\nWe completed a tour of the premises and saw that work had been completed to improve the environment \\nbut found that an upstairs window did not have a restrictor on it as required This meant it could be could be\\nopened fully which put people who used the service at risk\\n\\nThe service had no risk assessment regarding Legionnaires disease Maintenance personnel told us the \\nservice had hot and cold stored water in tanks which posed a potential risk of legionella they had not been \\ngiven any guidance regarding prevention of legionella A cold water tank at the top of the basement steps \\nwas in poor repair and leaking a plumber was called out to attend to the leak on the day of inspection \\nMaintenance personnel showed us water temperature checks they had completed these were to ensure the\\nhot water outlets were not above 43 degrees to prevent scalding There was no evidence that unused \\nshower heads were descaled disinfected and flushed They confirmed the hot water tank was set at 60 \\n\\n8 Amber House Inspection report 14 November 2017\\n\\n\\x0cdegrees centigrade which is required to help prevent legionella Senior staff told us they would contact an \\nappropriate company and ensure a risk assessment was completed for the risk of Legionnaires disease\\n\\nWe found items of fruit salad and vegetables that were past their recommended use by and best before \\ndates In freezers we found items of food which indicated on the packet that they could not be frozen There\\nwere also packets of unopened vegetables that were not sealed Items in the fridge such as processed meat \\nhad labels produced by catering staff of when these had been opened and when they should be used by \\nHowever we found the labels did not match the use before date once opened recommended by food \\nproducers on the packet These were mentioned to the registered manager to dispose of and address with \\ncatering staff\\n\\nThis demonstrated a continuing breach of Regulation 12 2 a b of the Health and Social Care Act 2008 \\nRegulated Activities Regulations 2014 safe care and treatment The action we have asked the provider to \\ntake can be found at the end of this report\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 12 2 h of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 safe care\\nand treatment We found concerns regarding infection prevention and control We saw there was no dirty to \\nclean flow within the laundry room Soiled washing was being stored in baskets on the floor alongside clean \\nlaundry also stored in open baskets creating the potential of cross contamination Areas of the service \\nrequired a deep clean and staff failed to use personal protective equipment PPE when required\\n\\nDuring this inspection we found that the environment had been refurbished to a high standard and the \\ncleanliness of the service had greatly improved Communal areas were clean and free from odours Rooms \\nhad been redecorated and bathrooms and toilets had been refitted We observed staff using PPE when \\nrequired and were told about the new laundry procedures in place We noted appropriate storage of dirty \\nand clean clothes in the laundry Mops and other cleaning products were stored appropriately and we saw \\nevidence that the domestic staffs weekly hours had increased by close to 50 All of these factors helped to \\nensure people were cared for in a clean and hygienic environment\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 13 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 safeguarding \\nservice users from abuse and improper treatment Staff had not always followed the providers safeguarding\\npolicies in practice For example the deputy manager began an investigation rather than consulting with \\nthe local safeguarding team to establish if this is what was required or whether the safeguarding team would\\ncomplete their own investigation We were told by staff and visiting professionals the training that staff \\nreceived on restrictive interventions known as BASIS training included details and a demonstration of \\nunapproved interventions that inflicted pain on the service users \\n\\nDuring this inspection we spoke with staff who had a clear understanding of the different types of abuse that\\nmay occur and the actions they would be required to take to keep people safe Records showed staff had \\ncompleted safeguarding vulnerable adults training A member of staff we spoke with said I would report \\nabuse to my manager immediately Its my duty\\n\\nThe consultant told us We are aware of the local authorities safeguarding procedures and would report \\nabuse to them but we may need to complete part of an investigation to establish some facts so we could be \\nassured our service users were safe\\n\\nWe saw evidence that staff had completed an accredited three day training course to ensure any use of \\n\\n9 Amber House Inspection report 14 November 2017\\n\\n\\x0cphysical interventions would be done safely This helped to reduce the risk that people would be injured \\nduring the delivery of essential care\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 18 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 staffing We \\nfound evidence that there had been several days when there was insufficient staff on duty to provide the \\nnecessary care and support to people This had impacted on the delivery of care for specific people who \\nused the service When we spoke with the registered manager they told us they did not use a dependency \\ntool to calculate staffing levels for the service They were unable to show us any systems they used to \\nestablish the amount of staff which should be on duty to ensure service users were supported on a one to \\none basis and other service users needs were effectively met\\n\\nDuring this comprehensive inspection we saw that staff were deployed in appropriate numbers to meet the \\nneeds of the people who used the service We saw evidence that a dependency tool was being utilised to \\nensure staffing levels were adequate At the time of the inspection there were eight people who lived in the \\nservice and one person who stayed for respite care four days a week Staff told us five people had one to one\\nsupport from staff for 12 hours each day from 8am to 8pm and one person had this level of support between\\n9am and 5pm A seventh person had one to one support for three hours a day which could be used flexibly \\nwhen required There were eight staff on duty during the day which included a senior care support worker \\nand four staff at night This level of staffing meant there was sufficient to meet the current numbers of \\npeople who used the service and the level of needs they had The registered manager was supernumerary to\\nthe rota and worked usual working hours during the week\\n\\nA member of staff commented Things used to get a bit crazy we would be running around and there never \\nseemed to be enough of us but things are different now I think it really helps them the people who used the\\nservice when we have more time to spend with them \\n\\nA visiting professional said Now they dont have the day service here things are much better When Ive \\nvisited I havent had concerns about the number of staff\\n\\n10 Amber House Inspection report 14 November 2017\\n\\n\\x0c Is the service effective\\n\\nRequires Improvement\\n\\nOur findings  \\nAt our comprehensive inspection in March and April 2017 we found the service to be noncompliant with \\nRegulation 11 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 need for \\nconsent We found the application of the Mental Capacity Act 2005 MCA was inconsistent People had \\nrestrictions in place around access to cigarettes and coffee at certain times Another person who used the \\nservice was having their mattress removed from their bedroom throughout the day We were unable to find \\nany documentation within the service to show the decision to remove the mattress had been discussed and \\nrecorded as in their best interest as the least restrictive option for them We also found staffs knowledge in \\nrelation to the MCA was limited\\n\\nDuring this inspection we found that improvements had been made in this area but further work was \\nrequired to ensure compliance with Regulation 11 The registered manager told us the restrictions around \\npeople coffee and cigarette intake continued and had been in place at a previous service They stated that it\\nwas in the persons best interest but there was no evidence available to support this The registered \\nmanager explained It is something that has always been in place and it does work really well but I will hold \\nmy hands up we havent had a best interest meeting about it The provider informed us that the person is \\nno longer living at the service so it no longer subjected to these restrictions \\n\\nFollowing the last inspection in March and April 2017 the registered manager told us that all staff would be \\nretrained in the MCA However at the time of this inspection no staff had completed refresher training in this\\narea\\n\\nThis demonstrated a continuing breach of Regulation 11 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 need for consent The action we have asked the provider to take can be found \\nat the end of this report\\n\\nWe saw evidence to confirm that best interest meetings had taken place and decisions were in place for \\nother aspects in the service including the removal of a persons mattress This had been agreed as the least \\nrestrictive option and necessary to ensure the person moved on in the daily routines A visiting \\nprofessional stated I have been involved in supporting the service so that appropriate and least restrictive \\ncare plans have been developed following best interest meetings\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 15 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 premises and \\nequipment We found concerns regarding the maintenance of the building A meter cupboard door was \\nfalling off its hinges when unlocked An unlocked cupboard was found to contain loose screws a knife an \\nopened paint can and other debris The unlocked cupboard had visible wires hanging and a light fitting \\nwithout a bulb connected to the electric supply We had issues with the lack of appropriate signage in the \\nkitchen accessed by people who used the service Radiators did not have covers in place as required When \\nwe requested a copy of the most recent electrical wiring certificate we found this was dated 1 December \\n2012 and valid for one year When we brought this to the attention of the registered manager they told us \\n\\n11 Amber House Inspection report 14 November 2017\\n\\n\\x0cthey would arrange to have this checked as a matter of urgency\\n\\nDuring this inspection we saw that the majority of issues identified at our previous inspection had been \\naddressed but further work was needed before compliance was achieved We saw evidence that an \\nelectrical wiring testing was complete in May 2017 Signage was in place cupboards in communal areas \\nwere locked and access to them was restricted However there was only one radiator cover in place and this\\nwas at the entrance to the service The registered manager explained The provider is making the radiator \\ncovers all the radiators are different sizes so each was is bespoke They are nearly finished and will be in \\nplace very soon\\n\\nWe found there had been a roof leak in a store room where care files were held These records had not been \\naffected but two boxes plastic stacking crates with lids containing personal items belonging to a person \\nwho used the service were wet and the contents at risk of damage These were moved There were other \\nissues such as wooden garden furniture which was water damaged that had left the wood split and pliable \\nThere were two wheel chairs in need of cleaning A team leader told us the garden furniture was to be \\ndisposed of and the wheelchairs were rarely used but they would see they were cleaned\\n\\nThis demonstrated a continuing breach of Regulation 15 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 premises and equipment The action we have asked the provider to take can be \\nfound at the end of this report\\n\\nDuring this inspection we saw records that confirmed staff had completed a range of training to equip them \\nwill the skills and abilities to complete their roles However a number of agency staff were being used at the \\nservice and their training records showed that had not completed any training in relation to supporting \\npeople with autism The consultant explained We do recognise that as an issue but there are no agencies \\nin the area that can supply us with staff who have that training They went on to say We do closely monitor\\nthe staff so we are satisfied they are supporting people in a person centred way\\n\\nThis demonstrated a continuing breach of Regulation 18 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 staffing The action we have asked the provider to take can be found at the end \\nof this report\\n\\nRecords showed staff received appropriate levels of support and supervision All staff had received a \\nsupervision session since the last inspection The supervision record had space to discuss the last \\nsupervision and any concerns or actions that had arisen from it Whether staff had any concerns and what \\nactions would be required to address them and any training or other recommendations \\n\\nPeoples health and social care needs were met by a range of healthcare professionals Over the course of \\nthe inspection a number of professionals visited the service A visiting professional told us I have worked \\nwith Amber House for years They have had issues in the past but I find they do everything they can for the \\npeople who live here I am always contacted if there is an issue and I feel that my advice is implemented and\\nwell received\\n\\nPeople were supported to eat and balanced diet of their choosing We saw the menus were written in \\nadvance and reflected the choices a younger age group would make The menus also included treats such \\nas a takeaway once a week We observed the lunchtime experience and saw that people were supported \\nappropriately and were provided with adapted beakers cutlery and plate guards to aid their independence\\n\\n12 Amber House Inspection report 14 November 2017\\n\\n\\x0c Is the service caring\\n\\nRequires Improvement\\n\\nOur findings  \\nWe saw care records and other personal or confidential information were stored securely Staff told us they \\nunderstood the importance of maintaining confidentiality Access to computerised records was restricted to\\nensure private and sensitive information could only be accessed by appropriate staff The registered \\nmanager explained We the management and administrative staff have our own passwords and different \\nlevels of access I can see things that other staff cant it depends what they need to know for their job \\n\\nWe asked to see records from the part of the service where supported living care was delivered The \\nregistered manager told us they did not have access to these records They said The senior staff that work \\nthere complete all of the paper work I dont really check it The consultant explained The records are held\\non a local compute which is not linked to Amber House This meant that the care and support of some \\npeople who used the service was not overseen by the registered manager as required Not having access to \\nthe information could lead to delays in specific care and support being delayed Following the inspection we\\nwere informed that the supported living part of the service would close to enable required improvements to \\ncontinue at Amber House\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 10 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 premises and \\nequipment While undertaking a tour of the building we observed a person who used the service naked \\ncompromising their dignity by undressing in public Staff had difficulty assisting and encouraging them to \\ndress in order to maintain their dignity We also heard staff asking a person if they needed to use the toilet in \\nfront of their peers \\n\\nDuring this comprehensive inspection we found the provider had ensured necessary improvements had \\nbeen made and compliance with this regulation had been achieved On a tour of the service we saw that \\ntowels were placed in strategic areas of the service so that if a person did remove their clothing staff could \\nprotect their modesty Care plans had been developed to ensure staff knew how to support people in these \\ncircumstances \\n\\nThroughout the inspection we saw staff treating people in a dignified and respectful way We did hear a \\nmember of staff speaking to a person in a way that did not appear age appropriate However when we \\nchecked their care plan we saw that this was a known strategy to reduce the persons anxiety A visiting \\nprofessional we spoke with confirmed this to be accurate \\n\\nWe did however see that a member of staff had recorded in the accident book that they had reprimanded \\na person who used the service following an incident of behaviour that challenged the service We \\nhighlighted this to the registered manager to address\\n\\nWe saw staff interacted well with people who used the service showing patience and demonstrating \\nknowledge of their preferences One of the people who used the service described the staff as Wonderful \\nAnother person said I can do what I want the staff are nice to me\\n\\n13 Amber House Inspection report 14 November 2017\\n\\n\\x0cPeople who used the service told us that they were encouraged to make decisions and choices in their daily \\nlives One person said I choose what clothes to wear and They make lovely meals I have a choice \\nAnother person said they had a nice room and had a DVD player so they could watch DVDs in their room\\n\\nWe saw evidence that people were given information and explanations in an appropriate format which \\npromoted their understanding Storyboards were used to support peoples understanding of specific \\nactivities such as appointments with healthcare professionals visits from family or events A storyboard uses\\npictorial symbols or photographs with numbers times and dates to enable people to understand what will \\noccur and when We also saw easy read versions of consent and capacity MCA and DoLS and No Secrets \\ndocuments in peoples care files\\n\\nWe saw that advocacy information was stored in peoples care files and that an advocacy group meeting \\nhad taken place The meeting was used to enable an independent mental capacity advocate IMCA to meet \\nthe people who used the service and start to develop relationships so people understood what support was \\navailable to them\\n\\n14 Amber House Inspection report 14 November 2017\\n\\n\\x0c Is the service responsive\\n\\nRequires Improvement\\n\\nOur findings  \\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 9 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 person centred \\ncare We found the quality of the instructions provided to staff to keep people safe was varied and brief For \\nexample one persons care plan said use distraction techniques but included no description of what these\\ndistraction techniques should be or for how long these should be used Other care plans described habitual \\nbehaviours but did not detail what these were or how staff could recognise them and stated provide me \\nwith my preferred method of  communication but did not describe or offer guidance as to what this was \\n\\nDuring this inspection we found that improvements had been made in this area but further work was \\nrequired to ensure compliance with regulation 9 The consultant told us We have had support from the \\nbehavioural support team but all of the care plans are live documents and a work in progress They will be \\nupdated as peoples needs changes or as the strategies we have put in place reduce some of their anxieties \\nand known behaviours\\n\\nWe saw that some peoples care plans included detailed and personalised guidance that enabled staff to \\nunderstand the support people required when they were showing signs of agitation However we found \\nother care plans still contained generic guidance For example one persons behavioural management plan \\nstated staff should use low arousal approach and distraction techniques and that a precursor to \\nbehaviours that challenged the service and others was indicated when facial expressions change Failing \\nto provide accurate and detailed guidance to staff means peoples needs would not be met consistently and\\neffectively\\n\\nDescriptions of when physical interventions were to be used were generic and open to interpretation One \\npersons behaviour management plan stated the approved techniques must be proportionate to the \\nbehaviour and applied for the minimum time necessary The descriptions of the support people required \\nwere open to interpretation and lacked clear guidance\\n\\nWe were told and saw evidence that care plans for people who lived at the service on a permanent basis had\\nbeen updated However the care plans for people who used the service on a respite basis still required re\\nwriting This meant there was a risk that staff did not have the guidance they required to deliver care in line \\nwith peoples needs and preferences \\n\\nThis demonstrated a continuing breach of Regulation 9 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 premises and equipment The action we have asked the provider to take can be \\nfound at the end of this report\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 16 1 2 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 receiving\\nand acting on complaints We spoke to relatives and relevant professionals who told us they had raised a \\nnumber of formal complaints to the registered manager These were in relation to the continuity of staff \\n\\n15 Amber House Inspection report 14 November 2017\\n\\n\\x0cstaffing levels the impact of day care service and subjective recording in daily records and not following \\ncare plans However the registered manager told us there had been no complaints since 2015\\n\\nDuring this inspection the consultant told us that improvements had been made with regards to the \\ncomplaints management within the service They said We have introduced a meet the manager day All \\nstakeholders service users relatives and professionals can meet the manager and discuss any issues or \\nthoughts they have It is just one way to show they are accessible and open to feedback\\n\\nThe providers complaints policy was displayed in the entrance to the service It had been produced in an \\neasy read format that helped to ensure it was accessible to the people who used the service Complaints \\ninformation was also available in peoples care plans\\n\\nA visiting relative we spoke with said I have had discussions with the manager in the past and because she \\nacted on them I didnt feel the need to raise anything formally I think the manager is receptive to feedback \\nand takes things on board A person who used the service said they had no complaints regarding the \\nservice they received They said None at all no reasons to\\n\\nDuring the inspection we saw people taking part in activities of their choosing One person told us they \\nenjoyed painting and arts and crafts Another person said they liked to go shopping and swimming A third \\nperson said I go to the shops to buy CDs DVDs and books\\n\\n16 Amber House Inspection report 14 November 2017\\n\\n\\x0c Is the service wellled\\n\\nInadequate\\n\\nOur findings  \\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulations 9 10 11 12 13 15 16 17 18 and 19 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 This meant there was also a breach of Regulation 8 General \\n\\nDuring this inspection we found that some improvements had been made across the service but further \\nwork was required before the service achieved compliance against regulations 9 11 12 15 17 18 and 19 \\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 17 of the Health and Social Care Act 2008 Regulated Activities Regulations 2014 good \\ngovernance There was no structured quality monitoring system in place that ensured identified shortfalls \\nwere addressed in a timely way \\n\\nDuring this inspection we saw that internal quality monitoring systems had not improved The consultant \\ntold us they were fully aware of the providers shortfalls in governance and would develop appropriate \\nsystems They said The governance arrangements across the organisation are being further developed to \\nreduce the sometimes disjointed approach and to enhance the collation interpretation and analysis of \\ninformation and subsequent action \\n\\nThe service did not have established and effective systems to assess monitor and improve the quality and \\nsafety of care or to assess and mitigate risks Accident and incident audits were not taking place the \\nregistered manager told us I do look at all of the incidents but I dont audit them Failing to review \\nincidents and identify patterns and trends could lead to preventable incidents taking place within the \\nservice \\n\\nCare plan audits were not being undertaken We saw that monthly care plan assessments were completed \\nbut these were not reviewed in conjunction with accident and incident records to establish in the guidance \\nprovided to staff remained appropriate They also failed to highlight that people were subject to restrictions \\nin their daily lives that had not been agreed as in their best interests\\n\\nThe provider had not undertaken audits of staff recruitment even though issues were highlighted at the last \\ninspection The provider had also failed to audit the documentation provided by the recruitment company \\nwho supplied agency staff This meant they had not assured themselves of the agency staffs suitability and \\nfitness to work with vulnerable people  \\n\\nWe saw that some medication audits had been completed on a monthly basis but these had not been \\neffective in highlighting that a person was not receiving their medicines as prescribed The in effective audits\\nbeing carried out lead to a person who had had been prescribed a particular medication to be taken every \\n72 hours receiving their medicines every 96 hours The shortfalls in auditing could have led to the person \\nexperiencing unnecessary pain \\n\\n17 Amber House Inspection report 14 November 2017\\n\\n\\x0cMedication audits had also not identified that some medicines had been written by the prescriber as \\ndirected which meant staff did not have adequate guidance to ensure staff knew have to administer \\nmedicines safely and consistently \\n\\nWater temperature checks were carried out by maintenance staff but there was no evidence to show that \\nlegionella checks were completed as required This meant risks to the health safety and welfare of the \\npeople who used the service were not being assessed or mitigated \\n\\nWe asked the registered manager to provide audits or the care and support provided in the supported living \\nservice The registered manager told us that they did not have managerial oversight and was not conducting\\nquality assurance checks they said We have a senior member of staff there but I have to hold my hands up\\nI dont monitor it Following the inspection the consultant informed us that the provider intended to de\\nregister the supported living service This would be done following a notice period being served to the \\npeople who used that part of the service\\n\\nThere were no systems and processes to ensure the safe storage and disposal of food items We identified \\nconcerns with out of date past the used by date and nonfreeze food items When we highlighted our \\nconcerns immediate action was taken but the provider had no systems to identify issues with food storage \\nand safety\\n\\nAt our comprehensive inspection in March and April 2017 and the subsequent report we highlighted \\nconcerns that radiators did not have covers and posed a risk to the people who used the service Despite \\nthis during this inspection we found only one radiator cover was in place This showed that provider did not \\nhave effective systems in place to mitigate known risks \\n\\nThe consultant told us We are completely in a limbo situation we have actions plans so know the areas \\nthat we need to improve We arent auditing the work we have just completed because we know its been \\ndone to a high standard and we are not auditing the other areas because we know that needs to be \\nupdated Any audits we did now would just prevent us fixing us things\\n\\nThis demonstrated a continuing breach of Regulation 17 of the Health and Social Care Act 2008 Regulated \\nActivities Regulations 2014 need for good governance The action we have asked the provider to take can \\nbe found at the end of this report\\n\\nAt our comprehensive inspection in March and April 2017 we found the service to be noncomplaint with \\nRegulation 18 of the Health and Social Care Act 2008 registration Regulations 2009 notification of other \\nincidents The registered manager failed to notify the Commission of two incidents where people had \\ninjured themselves and required hospital treatment \\n\\nDuring this inspection we reviewed the accident and incident records held at the service against the \\ninformation we had received from the provider We found that we had been informed of all notifiable events \\nas required The registered manager commented The incidents that didnt get reported were a mistake I \\nknow what needs to be reported We try to be open and transparent about everything that happens\\n\\nThe registered manager told us they had not been supported in their role to ensure positive outcomes for \\npeople They said The directors are very supportive in some areas but they dont have a background in \\ncare so when it came down to the residents I was left to get on with things myself I have had more support \\nfrom name of the consultant in the short time they have been here than I ever have before The consultant\\ntold us We know there are a lot of improvements to make I initially came on board for nine weeks and that \\n\\n18 Amber House Inspection report 14 November 2017\\n\\n\\x0chas already been extended to 12 months I am committed to turning things around and making this a good \\nservice for all of our residents\\n\\n19 Amber House Inspection report 14 November 2017\\n\\n\\x0cThis section is primarily information for the provider\\n\\nAction we have told the provider to take\\n\\nThe table below shows where regulations were not being met and we have asked the provider to send us a \\nreport that says what action they are going to takeWe will check that this action is taken by the provider\\n\\nRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nRegulation\\nRegulation 9 HSCA RA Regulations 2014 Person\\ncentred care\\n\\nPersonal care\\n\\nPeoples care plans did not always reflect their \\ncurrent needs and were not contain \\nappropriate information to ensure their needs \\nwould be met\\n\\nRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nRegulation\\nRegulation 11 HSCA RA Regulations 2014 Need \\nfor consent\\n\\nPersonal care\\n\\nPeople who used the service were subjected to \\nrestrictions that had not been agreed as in their\\nbest interests\\n\\nRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nRegulation\\nRegulation 12 HSCA RA Regulations 2014 Safe \\ncare and treatment\\n\\nPersonal care\\n\\nPeople did not receive their medicines as \\nprescribed \\n\\nThe provider had not taken appropriate action \\nto mitigate known risks as required\\n\\nRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nRegulation\\nRegulation 15 HSCA RA Regulations 2014 \\nPremises and equipment\\n\\nPersonal care\\n\\nThe registered provider had failed to ensure the\\npremises were appropriate and adequately \\nmaintained\\n\\n20 Amber House Inspection report 14 November 2017\\n\\n\\x0cRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nRegulation\\nRegulation 17 HSCA RA Regulations 2014 Good \\ngovernance\\n\\nPersonal care\\n\\nRegulated activity\\nAccommodation for persons who require nursing or \\npersonal care\\n\\nPersonal care\\n\\nThe providers governance systems did not \\ncover all aspects of the service and were \\nineffective at highlighting shortfalls\\n\\nRegulation\\nRegulation 18 HSCA RA Regulations 2014 Staffing\\n\\nStaff had not always completed training to \\nensure they had the necessary skills to meet the\\nindividual needs of the people who used the \\nservice\\n\\n21 Amber House Inspection report 14 November 2017\\n\\n\\x0c'"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unpunkt(text_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### N-Grams as Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['this', 'is'], ['is', 'a'], ['a', 'sample'], ['sample', 'text']]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "def generate_ngrams(text, n):\n",
    "    words = text.split()\n",
    "    output = []  \n",
    "    for i in range(len(words)-n+1):\n",
    "        output.append(words[i:i+n])\n",
    "    return output\n",
    "\n",
    "# Binary gram \n",
    "generate_ngrams('this is a sample text', 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk import bigrams, trigrams\n",
    "from nltk import word_tokenize\n",
    "\n",
    "sent = word_tokenize(\"The quick brown fox jumps over the lazy dog\")\n",
    "sent_bigrams = bigrams(sent)\n",
    "sent_trigrams = trigrams(sent)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('The', 'quick'),\n",
       " ('quick', 'brown'),\n",
       " ('brown', 'fox'),\n",
       " ('fox', 'jumps'),\n",
       " ('jumps', 'over'),\n",
       " ('over', 'the'),\n",
       " ('the', 'lazy'),\n",
       " ('lazy', 'dog')]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(sent_bigrams)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('The', 'quick', 'brown'),\n",
       " ('quick', 'brown', 'fox'),\n",
       " ('brown', 'fox', 'jumps'),\n",
       " ('fox', 'jumps', 'over'),\n",
       " ('jumps', 'over', 'the'),\n",
       " ('over', 'the', 'lazy'),\n",
       " ('the', 'lazy', 'dog')]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(sent_trigrams)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Regular expressions\n",
    "Regular expressions (called REs, or regexes, or regex patterns) are essentially a tiny, \n",
    "highly specialized programming language embedded inside Python and made available through the re module.\n",
    "Using this little language, you specify the rules for the set of possible strings that you want to match; \n",
    "this set might contain English sentences, or e-mail addresses, or TeX commands, or anything you like. \n",
    "You can then ask questions such as:\n",
    "“Does this string match the pattern?”, or “Is there a match for the pattern anywhere in this string?”. \n",
    "You can also use REs to modify a string or to split it apart in various ways.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'remove this  from this website'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sample code to remove a regex pattern \n",
    "import re\n",
    "\n",
    "\n",
    "def _remove_regex(input_text, regex_pattern):\n",
    "    urls = re.finditer(regex_pattern, input_text) \n",
    "    for i in urls: \n",
    "        input_text = re.sub(i.group().strip(), '', input_text)\n",
    "    return input_text\n",
    "\n",
    "regex_pattern = \"#[\\w]*\" \n",
    "\n",
    "_remove_regex(\"remove this #hashtag from this website\", regex_pattern)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#Using this little language, you specify the rules for the set of possible strings that you want to match;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['data science',\n",
       " 'Data Scientist',\n",
       " 'data science',\n",
       " 'data scientist',\n",
       " 'Data science',\n",
       " 'data science']"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "htmltext ='''   <div class=\"container\">\n",
    "      <h3 class=\"quotes__subtitle chapeau-title chapeau-title--primary\">Testimonials</h3>\n",
    "      <h2 class=\"u-text-center\">Don’t just take our word for it.</h2>\n",
    "    <div class=\"quotes__wrapper js-quotes-height\">\n",
    "      <div class=\"js-quote\" data-index=\"0\" style=\"\">\n",
    "        <div class=\"quotes__quote\">The courses are excellent—they meet a higher standard than other platforms.</div>\n",
    "        <div class=\"quotes__description\">Helge is a PhD. Candidate in sociology and is learning data science tools to augment his research.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"1\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">Every month I learn something new on DataCamp, and they are all things that are very useful in my job.</div>\n",
    "        <div class=\"quotes__description\">Godefroy Clair is a Data Scientist at Flylab, where he helps create software for autonomous drones.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"2\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">DataCamp was essential for getting my job, which was my break into data science.</div>\n",
    "        <div class=\"quotes__description\">Jamen Long used the skills he learned at DataCamp to land a full time job as a data scientist.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"3\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">Data science is the future, and it is better to be on the cutting edge than left behind.</div>\n",
    "        <div class=\"quotes__description\">Arnaud uses techniques he&#39;s learned on DataCamp every day in his work for the French government.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"4\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">If anything, I regret I didn&#39;t start with DataCamp earlier.</div>\n",
    "        <div class=\"quotes__description\">Brigham Young University uses DataCamp to help students grasp data science concepts.</div>\n",
    "      </div>\n",
    "    </div>\n",
    "    '''\n",
    "\n",
    "\n",
    "# i want to find all the strings that:\n",
    "# start either with capital or lower d then include 'ata' then have space and then they consist of a word\n",
    "# i describe this pattern as follows:\n",
    "re.findall(r'[dD]ata [A-Za-z]+',htmltext)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#You can also use Regular expressions to modify a string\n",
    "def preprocesshtml(text):\n",
    "    # convert to ASCII\n",
    "\n",
    "    # if the input is HTML, force-add full stops after these tags\n",
    "    fullStopTags = ['li', 'p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'dd']\n",
    "    for tag in fullStopTags:\n",
    "        text = re.sub(r'</'+tag+'>', '.', text)\n",
    "    text = re.sub(r'<[^>]+>', '', text)                  # strip out HTML\n",
    "    text = re.sub(r'[,:;()\\-]', ' ', text)               # replace commas, hyphens etc (count as spaces)\n",
    "    text = re.sub(r'[\\.!?]', '.', text)                  # unify terminators\n",
    "    text = re.sub(r'^\\s+', '', text)                     # strip leading whitespace\n",
    "    text = re.sub(r'[ ]*(\\n|\\r\\n|\\r)[ ]*', ' ', text)    # replace new lines with spaces\n",
    "    text = re.sub(r'([\\.])[\\. ]+', '.', text)            # check for duplicated terminators\n",
    "    text = re.sub(r'[ ]*([\\.])', '. ', text)             # pad sentence terminators\n",
    "    text = re.sub(r'\\s+', ' ', text)                     # remove multiple spaces\n",
    "    text = re.sub(r'\\s+$', '', text);                    # strip trailing whitespace\n",
    "    return text\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "htmltext ='''   <div class=\"container\">\n",
    "      <h3 class=\"quotes__subtitle chapeau-title chapeau-title--primary\">Testimonials</h3>\n",
    "      <h2 class=\"u-text-center\">Don’t just take our word for it.</h2>\n",
    "    <div class=\"quotes__wrapper js-quotes-height\">\n",
    "      <div class=\"js-quote\" data-index=\"0\" style=\"\">\n",
    "        <div class=\"quotes__quote\">The courses are excellent—they meet a higher standard than other platforms.</div>\n",
    "        <div class=\"quotes__description\">Helge is a PhD. Candidate in sociology and is learning data science tools to augment his research.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"1\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">Every month I learn something new on DataCamp, and they are all things that are very useful in my job.</div>\n",
    "        <div class=\"quotes__description\">Godefroy Clair is a Data Scientist at Flylab, where he helps create software for autonomous drones.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"2\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">DataCamp was essential for getting my job, which was my break into data science.</div>\n",
    "        <div class=\"quotes__description\">Jamen Long used the skills he learned at DataCamp to land a full time job as a data scientist.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"3\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">Data science is the future, and it is better to be on the cutting edge than left behind.</div>\n",
    "        <div class=\"quotes__description\">Arnaud uses techniques he&#39;s learned on DataCamp every day in his work for the French government.</div>\n",
    "      </div>\n",
    "      <div class=\"js-quote\" data-index=\"4\" style=\"display: none;\">\n",
    "        <div class=\"quotes__quote\">If anything, I regret I didn&#39;t start with DataCamp earlier.</div>\n",
    "        <div class=\"quotes__description\">Brigham Young University uses DataCamp to help students grasp data science concepts.</div>\n",
    "      </div>\n",
    "    </div>\n",
    "    '''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Testimonials. Don’t just take our word for it. The courses are excellent—they meet a higher standard than other platforms. Helge is a PhD. Candidate in sociology and is learning data science tools to augment his research. Every month I learn something new on DataCamp and they are all things that are very useful in my job. Godefroy Clair is a Data Scientist at Flylab where he helps create software for autonomous drones. DataCamp was essential for getting my job which was my break into data science. Jamen Long used the skills he learned at DataCamp to land a full time job as a data scientist. Data science is the future and it is better to be on the cutting edge than left behind. Arnaud uses techniques he&#39 s learned on DataCamp every day in his work for the French government. If anything I regret I didn&#39 t start with DataCamp earlier. Brigham Young University uses DataCamp to help students grasp data science concepts.'"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocesshtml(htmltext)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Testimonials Don’t just take our word for it The courses are excellentthey meet a higher standard than other platforms Helge is a PhD Candidate in sociology and is learning data science tools to augment his research Every month I learn something new on DataCamp and they are all things that are very useful in my job Godefroy Clair is a Data Scientist at Flylab where he helps create software for autonomous drones DataCamp was essential for getting my job which was my break into data science Jamen Long used the skills he learned at DataCamp to land a full time job as a data scientist Data science is the future and it is better to be on the cutting edge than left behind Arnaud uses techniques he39 s learned on DataCamp every day in his work for the French government If anything I regret I didn39 t start with DataCamp earlier Brigham Young University uses DataCamp to help students grasp data science concepts'"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unpunkt(preprocesshtml(htmltext))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Lexicon Normalization\n",
    "\n",
    "\n",
    "Normalization is a pivotal step for feature engineering with text as it converts the high dimensional features (N different features) to the low dimensional space (1 feature), which is an ideal ask for any ML model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "#### Stemming\n",
    "\n",
    "Stemming is a rudimentary rule-based process of stripping the suffixes (“ing”, “ly”, “es”, “s” etc) from a word."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "multipli\n",
      "integr\n"
     ]
    }
   ],
   "source": [
    "from nltk.stem.porter import PorterStemmer \n",
    "\n",
    "# Defining objects\n",
    "stem = PorterStemmer()\n",
    "\n",
    "word_1 = \"multiplying\" \n",
    "word_2 = \"integrating\"\n",
    "\n",
    "print(stem.stem(word_1))\n",
    "print(stem.stem(word_2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "#### Lemmatization\n",
    "\n",
    "Lemmatization, on the other hand, is an organized & step by step procedure of obtaining the root form of the word, it makes use of vocabulary (dictionary importance of words) and morphological analysis (word structure and grammar relations)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "multiply\n",
      "integrate\n"
     ]
    }
   ],
   "source": [
    "from nltk import WordNetLemmatizer\n",
    "\n",
    "# Defining objects \n",
    "lem = WordNetLemmatizer()\n",
    "\n",
    "word_1 = \"multiplying\" \n",
    "word_2 = \"integrating\"\n",
    "\n",
    "print(lem.lemmatize(word_1, \"v\"))\n",
    "print(lem.lemmatize(word_2, \"v\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### When we sent_tokenize a string we produce a list (of sentences).\n",
    "### Then from that we can word_tokenize a list of words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.6/site-packages/nltk/twitter/__init__.py:20: UserWarning: The twython library has not been installed. Some functionality from the twitter package will not be available.\n",
      "  warnings.warn(\"The twython library has not been installed. \"\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from operator import itemgetter\n",
    "import string\n",
    "import numpy as np\n",
    "from collections import Counter\n",
    "\n",
    "from nltk.sentiment.vader import SentimentIntensityAnalyzer\n",
    "from nltk import word_tokenize\n",
    "from nltk.corpus import wordnet as wn\n",
    "from nltk import wordpunct_tokenize\n",
    "from nltk import WordNetLemmatizer\n",
    "from nltk import sent_tokenize\n",
    "from nltk import pos_tag\n",
    "from nltk.corpus import stopwords \n",
    "\n",
    "from nltk.corpus import wordnet\n",
    "\n",
    "\n",
    "wordnet_lemmatizer = WordNetLemmatizer()\n",
    "\n",
    "def tokenize(document):\n",
    "        \"\"\"\n",
    "        Returns a normalized, lemmatized list of tokens from a document by\n",
    "        applying segmentation (breaking into sentences), then word/punctuation\n",
    "        tokenization, and finally part of speech tagging. It uses the part of\n",
    "        speech tags to look up the lemma in WordNet, and returns the lowercase\n",
    "        version of all the words, removing stopwords and punctuation.\n",
    "        \"\"\"\n",
    "        tokens=[]\n",
    "        \n",
    "        \n",
    "        # Break the document into sentences\n",
    "        for sent in sent_tokenize(document):\n",
    "            # Break the sentence into part of speech tagged tokens\n",
    "            for token, tag in pos_tag(wordpunct_tokenize(sent)):\n",
    "\n",
    "\n",
    "                # If  stopword, ignore token and continue otherwise provide lemma\n",
    "                if token in stopWords:\n",
    "                    continue\n",
    "                else:\n",
    "                    tokens.append(wordnet_lemmatizer.lemmatize(token).lower())\n",
    "                \n",
    "            \n",
    "        return tokens\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['wordnet',\n",
       " 'large',\n",
       " 'lexical',\n",
       " 'database',\n",
       " 'english',\n",
       " 'nouns',\n",
       " 'verb',\n",
       " 'adjective',\n",
       " 'adverb',\n",
       " 'grouped',\n",
       " 'set',\n",
       " 'cognitive',\n",
       " 'synonym',\n",
       " 'synset',\n",
       " 'expressing',\n",
       " 'distinct',\n",
       " 'concept',\n",
       " 'synsets',\n",
       " 'interlinked',\n",
       " 'mean',\n",
       " 'conceptualsemantic',\n",
       " 'lexical',\n",
       " 'relation',\n",
       " 'the',\n",
       " 'resulting',\n",
       " 'network',\n",
       " 'meaningfully',\n",
       " 'related',\n",
       " 'word',\n",
       " 'concept',\n",
       " 'navigated',\n",
       " 'browser',\n",
       " 'wordnet',\n",
       " 'also',\n",
       " 'freely',\n",
       " 'publicly',\n",
       " 'available',\n",
       " 'download',\n",
       " 'wordnet',\n",
       " '’',\n",
       " 'structure',\n",
       " 'make',\n",
       " 'useful',\n",
       " 'tool',\n",
       " 'computational',\n",
       " 'linguistics',\n",
       " 'natural',\n",
       " 'language',\n",
       " 'processing',\n",
       " 'wordnet',\n",
       " 'superficially',\n",
       " 'resembles',\n",
       " 'thesaurus',\n",
       " 'group',\n",
       " 'word',\n",
       " 'together',\n",
       " 'based',\n",
       " 'meaning',\n",
       " 'however',\n",
       " 'important',\n",
       " 'distinction',\n",
       " 'first',\n",
       " 'wordnet',\n",
       " 'interlinks',\n",
       " 'word',\n",
       " 'formsstrings',\n",
       " 'lettersbut',\n",
       " 'specific',\n",
       " 'sens',\n",
       " 'word',\n",
       " 'as',\n",
       " 'result',\n",
       " 'word',\n",
       " 'found',\n",
       " 'close',\n",
       " 'proximity',\n",
       " 'one',\n",
       " 'another',\n",
       " 'network',\n",
       " 'semantically',\n",
       " 'disambiguated',\n",
       " 'second',\n",
       " 'wordnet',\n",
       " 'label',\n",
       " 'semantic',\n",
       " 'relation',\n",
       " 'among',\n",
       " 'word',\n",
       " 'whereas',\n",
       " 'grouping',\n",
       " 'word',\n",
       " 'thesaurus',\n",
       " 'follow',\n",
       " 'explicit',\n",
       " 'pattern',\n",
       " 'meaning',\n",
       " 'similarity']"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text='''     WordNet is a large lexical database of English. Nouns, verbs, adjectives and adverbs are grouped \n",
    "into sets of cognitive synonyms (synsets), each expressing a distinct concept. \n",
    "Synsets are interlinked by means of conceptual-semantic and lexical relations. \n",
    "The resulting network of meaningfully related words and concepts can be navigated with the browser. \n",
    "WordNet is also freely and publicly available for download. WordNet’s structure makes it \n",
    " a useful tool for computational linguistics and natural language processing.\n",
    "\n",
    "WordNet superficially resembles a thesaurus, in that it groups words together based on their meanings. \n",
    "However, there are some important distinctions. First, WordNet interlinks not just word forms—strings \n",
    "of letters—but specific senses of words. As a result, words that are found in close proximity to one \n",
    "another in the network are semantically disambiguated. Second, WordNet labels the semantic \n",
    "relations among words, whereas the groupings of words in a thesaurus does not \n",
    "follow any explicit pattern other than meaning similarity '''\n",
    "\n",
    "\n",
    "\n",
    "(tokenize(unpunkt(text)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['testimonials',\n",
       " 'don',\n",
       " '’',\n",
       " 'take',\n",
       " 'word',\n",
       " 'the',\n",
       " 'course',\n",
       " 'excellentthey',\n",
       " 'meet',\n",
       " 'higher',\n",
       " 'standard',\n",
       " 'platform',\n",
       " 'helge',\n",
       " 'phd',\n",
       " 'candidate',\n",
       " 'sociology',\n",
       " 'learning',\n",
       " 'data',\n",
       " 'science',\n",
       " 'tool',\n",
       " 'augment',\n",
       " 'research',\n",
       " 'every',\n",
       " 'month',\n",
       " 'i',\n",
       " 'learn',\n",
       " 'something',\n",
       " 'new',\n",
       " 'datacamp',\n",
       " 'thing',\n",
       " 'useful',\n",
       " 'job',\n",
       " 'godefroy',\n",
       " 'clair',\n",
       " 'data',\n",
       " 'scientist',\n",
       " 'flylab',\n",
       " 'help',\n",
       " 'create',\n",
       " 'software',\n",
       " 'autonomous',\n",
       " 'drone',\n",
       " 'datacamp',\n",
       " 'essential',\n",
       " 'getting',\n",
       " 'job',\n",
       " 'break',\n",
       " 'data',\n",
       " 'science',\n",
       " 'jamen',\n",
       " 'long',\n",
       " 'used',\n",
       " 'skill',\n",
       " 'learned',\n",
       " 'datacamp',\n",
       " 'land',\n",
       " 'full',\n",
       " 'time',\n",
       " 'job',\n",
       " 'data',\n",
       " 'scientist',\n",
       " 'data',\n",
       " 'science',\n",
       " 'future',\n",
       " 'better',\n",
       " 'cutting',\n",
       " 'edge',\n",
       " 'left',\n",
       " 'behind',\n",
       " 'arnaud',\n",
       " 'us',\n",
       " 'technique',\n",
       " 'he39',\n",
       " 'learned',\n",
       " 'datacamp',\n",
       " 'every',\n",
       " 'day',\n",
       " 'work',\n",
       " 'french',\n",
       " 'government',\n",
       " 'if',\n",
       " 'anything',\n",
       " 'i',\n",
       " 'regret',\n",
       " 'i',\n",
       " 'didn39',\n",
       " 'start',\n",
       " 'datacamp',\n",
       " 'earlier',\n",
       " 'brigham',\n",
       " 'young',\n",
       " 'university',\n",
       " 'us',\n",
       " 'datacamp',\n",
       " 'help',\n",
       " 'student',\n",
       " 'grasp',\n",
       " 'data',\n",
       " 'science',\n",
       " 'concept']"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenize(unpunkt(preprocesshtml(htmltext)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "we can tokenise strings in dataframe using df.apply\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    [The, service, was, effective, ., There, were,...\n",
      "1    [The, service, was, effective, ., People, 's, ...\n",
      "2    [The, service, was, not, consistently, effecti...\n",
      "3    [The, service, was, not, always, effective, .,...\n",
      "4    [The, service, was, effective, ., People, rece...\n",
      "Name: tokenized_sents, dtype: object\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import nltk\n",
    "\n",
    "\n",
    "df['tokenized_sents'] = df.apply(lambda row: nltk.word_tokenize(row[5]), axis=1)\n",
    "print(df['tokenized_sents'].head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    0             outstanding\n",
      "1             outsta...\n",
      "1    0    miss s g howard victoria lodge care home ...\n",
      "2    0      is the service safe?  is the service ef...\n",
      "3    0      this unannounced inspection took place ...\n",
      "4    0      the service was safe  people told us th...\n",
      "dtype: object\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0     0             outstanding\n",
       "1             outsta...\n",
       "1     0    miss s g howard victoria lodge care home ...\n",
       "2     0      is the service safe?  is the service ef...\n",
       "3     0      this unannounced inspection took place ...\n",
       "4     0      the service was safe  people told us th...\n",
       "5     0      the service was effective.   there were...\n",
       "6     0      the service was very caring.  people re...\n",
       "7     0      the service was extremely responsive.  ...\n",
       "8     0      the service was very well led.  people ...\n",
       "9     0     we carried out this inspection under sec...\n",
       "10    0                      good  \n",
       "1               ...\n",
       "11    0       people and relatives all gave us very ...\n",
       "12    0                      good  \n",
       "1               ...\n",
       "13    0       people and relatives we spoke with tol...\n",
       "14    0               outstanding  \n",
       "1               ...\n",
       "15    0       people living in victoria lodge were h...\n",
       "16    0                      good  \n",
       "1               ...\n",
       "17    0       relatives told us that the supportive ...\n",
       "18    0               outstanding  \n",
       "1               ...\n",
       "19    0       the registered manager and provider pr...\n",
       "20    0       the registered manager and provider pr...\n",
       "dtype: object"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import nltk\n",
    "\n",
    "\n",
    "# df1 = df.apply(lambda x: x[0].lower(), axis=1)\n",
    "# print(df1.head())\n",
    "\n",
    "\n",
    "# columns = [0,21]\n",
    "for column in range(0,21) :\n",
    "    df1[column] = df[column].apply(lambda x: x.lower()).copy()\n",
    "    \n",
    "print(df1.head())\n",
    "\n",
    "df1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Object Standardization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'This is a Retweet retweeted tweet by me and its awesome'"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text_raw =\"This is a rt retweeted tweet by me and its awsm\"\n",
    "\n",
    "lookup_dict = {'rt':'Retweet', 'dm':'direct message', \"awsm\" : \"awesome\", \"luv\" :\"love\"}\n",
    "\n",
    "def _lookup_words(input_text):\n",
    "    words = input_text.split() \n",
    "    new_words = [] \n",
    "    for word in words:\n",
    "        if word.lower() in lookup_dict:\n",
    "            word = lookup_dict[word.lower()]\n",
    "        new_words.append(word) \n",
    "        new_text = \" \".join(new_words) \n",
    "    return new_text\n",
    "\n",
    "_lookup_words(text_raw)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "#### Text to Features\n",
    "#### Syntactic Parsing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('I', 'PRP'), ('am', 'VBP'), ('learning', 'VBG'), ('Natural', 'NNP'), ('Language', 'NNP'), ('Processing', 'NNP'), ('on', 'IN'), ('Analytics', 'NNP'), ('Vidhya', 'NNP')]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('Please', 'NNP'),\n",
       " ('book', 'NN'),\n",
       " ('my', 'PRP$'),\n",
       " ('flight', 'NN'),\n",
       " ('for', 'IN'),\n",
       " ('Delhi', 'NNP')]"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk import word_tokenize, pos_tag\n",
    "\n",
    "text = \"I am learning Natural Language Processing on Analytics Vidhya\"\n",
    "tokens = word_tokenize(text)\n",
    "\n",
    "print(pos_tag(tokens))\n",
    "\n",
    "## Proper Noun:\n",
    "#NNP (Proper Noun)\n",
    "# VBG Verb gerund \n",
    "\n",
    "setence_1 = \"Please book my flight for Delhi\"\n",
    "\n",
    "setence_2 = \"I am going to read this book in the flight\"\n",
    "\n",
    "\n",
    "# Context based ?\n",
    "\n",
    "pos_tag(word_tokenize(setence_2))\n",
    "\n",
    "pos_tag(word_tokenize(setence_1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### The process of detecting the named entities such as person names, location names, company names etc from the text is called as NER."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Data Analysis\n",
    "\n",
    "Now that we have cleaned our text and formatted the data structure, we can carry out some analysis - *yipee!*\n",
    "\n",
    "---\n",
    "#### Topic Modelling \n",
    "\n",
    "Topic modeling is a process of automatically identifying the topics present in a text corpus, it derives the hidden patterns among the words in the corpus in an unsupervised manner. \n",
    "\n",
    ">Topics are defined as “a repeating pattern of co-occurring terms in a corpus”. \n",
    "\n",
    "> A good topic model results in – “health”, “doctor”, “patient”, “hospital” for a topic – Healthcare, and “farm”, “crops”, “wheat” for a topic – “Farming”."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "#### Latent Dirichlet Allocation (LDA)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['Sugar',\n",
       "  'is',\n",
       "  'bad',\n",
       "  'to',\n",
       "  'consume.',\n",
       "  'My',\n",
       "  'sister',\n",
       "  'likes',\n",
       "  'to',\n",
       "  'have',\n",
       "  'sugar,',\n",
       "  'but',\n",
       "  'not',\n",
       "  'my',\n",
       "  'father.'],\n",
       " ['My',\n",
       "  'father',\n",
       "  'spends',\n",
       "  'a',\n",
       "  'lot',\n",
       "  'of',\n",
       "  'time',\n",
       "  'driving',\n",
       "  'my',\n",
       "  'sister',\n",
       "  'around',\n",
       "  'to',\n",
       "  'dance',\n",
       "  'practice.'],\n",
       " ['Doctors',\n",
       "  'suggest',\n",
       "  'that',\n",
       "  'driving',\n",
       "  'may',\n",
       "  'cause',\n",
       "  'increased',\n",
       "  'stress',\n",
       "  'and',\n",
       "  'blood',\n",
       "  'pressure.']]"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "doc1 = \"Sugar is bad to consume. My sister likes to have sugar, but not my father.\" \n",
    "doc2 = \"My father spends a lot of time driving my sister around to dance practice.\"\n",
    "doc3 = \"Doctors suggest that driving may cause increased stress and blood pressure.\"\n",
    "\n",
    "doc_complete = [doc1, doc2, doc3]\n",
    "doc_clean = [doc.split() for doc in doc_complete]\n",
    "\n",
    "doc_clean"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Gensim\n",
    "\n",
    "> Gensim is a free Python library designed to automatically extract semantic topics from documents, as efficiently (computer-wise) and painlessly (human-wise) as possible. https://radimrehurek.com/gensim/index.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'gensim'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-54-93c367e0ff36>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mgensim\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mgensim\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mcorpora\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'gensim'"
     ]
    }
   ],
   "source": [
    "import gensim\n",
    "\n",
    "from gensim import corpora"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'corpora' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-55-2ac7d7ae880a>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# Creating the term dictionary of our corpus, where every unique term is assigned an index.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mdictionary\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcorpora\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDictionary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdoc_clean\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;31m#Converting list of documents (corpus) into Document\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;31m#Term Matrix using dictionary prepared above.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'corpora' is not defined"
     ]
    }
   ],
   "source": [
    "# Creating the term dictionary of our corpus, where every unique term is assigned an index.  \n",
    "dictionary = corpora.Dictionary(doc_clean)\n",
    "\n",
    "#Converting list of documents (corpus) into Document\n",
    "#Term Matrix using dictionary prepared above. \n",
    "## ?? vectors number of times each elements appears\n",
    "doc_term_matrix = [dictionary.doc2bow(doc) for doc in doc_clean]\n",
    "\n",
    "print (doc_term_matrix)\n",
    "\n",
    "# Creating the object for LDA model using gensim library\n",
    "Lda = gensim.models.ldamodel.LdaModel\n",
    "\n",
    "# Running and Training LDA model on the document term matrix\n",
    "ldamodel = Lda(doc_term_matrix, num_topics=3, id2word = dictionary, passes=50)\n",
    "\n",
    "# Results \n",
    "print(ldamodel.print_topics())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'corpora' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-56-df7b55f73d7e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mdoc_clean\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mtext\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mdictionary\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcorpora\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDictionary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdoc_clean\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mdoc_term_matrix\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mdictionary\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdoc2bow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdoc\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mdoc\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mdoc_clean\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'corpora' is not defined"
     ]
    }
   ],
   "source": [
    "text=\"He is a big big boy\"\n",
    "\n",
    "doc_clean= [text.split()]\n",
    "dictionary = corpora.Dictionary(doc_clean)\n",
    "doc_term_matrix = [dictionary.doc2bow(doc) for doc in doc_clean]\n",
    "\n",
    "# Running and Training LDA model on the document term matrix\n",
    "ldamodel = Lda(doc_term_matrix, num_topics=3, id2word = dictionary, passes=50)\n",
    "\n",
    "# Results \n",
    "print(ldamodel.print_topics()[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Word Embedding (text vectors)\n",
    "\n",
    "#### Word2Vec\n",
    "#### GloVe \n",
    "\n",
    "Word2Vec model is composed of preprocessing module, a shallow neural network model called Continuous Bag of Words and another shallow neural network model called skip-gram. These models are widely used for all other nlp problems. It first constructs a vocabulary from the training corpus and then learns word embedding representations. Following code using gensim package prepares the word embedding as the vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'gensim'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-59-b6db5230ad79>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mgensim\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodels\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mWord2Vec\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m sentences = [['data', 'science'], ['vidhya', 'science', 'data', 'analytics'],\n\u001b[1;32m      4\u001b[0m              ['machine', 'learning'], ['deep', 'learning']]\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'gensim'"
     ]
    }
   ],
   "source": [
    "from gensim.models import Word2Vec\n",
    "\n",
    "sentences = [['data', 'science'], ['vidhya', 'science', 'data', 'analytics'],\n",
    "             ['machine', 'learning'], ['deep', 'learning']]\n",
    "\n",
    "\n",
    "# train the model on your corpus\n",
    "model = Word2Vec(sentences, min_count = 1)\n",
    "\n",
    "print(model.similarity('data', 'science'))\n",
    "\n",
    "print(model['learning']) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Bag-of-Words Model\n",
    "\n",
    "We cannot work with text directly when using machine learning algorithms.\n",
    "\n",
    "Instead, we need to convert the text to numbers.\n",
    "\n",
    "We may want to perform classification of documents, so each document is an “input” and a class label is the “output” for our predictive algorithm. Algorithms take vectors of numbers as input, therefore we need to convert documents to fixed-length vectors of numbers.\n",
    "\n",
    "A simple and effective model for thinking about text documents in machine learning is called the Bag-of-Words Model, or BoW.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>the cat with the hat sat in the mat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>you have brains in your head</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>you have feet in your shoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>You can steer yourself any direction you choose</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Think and wonder, wonder and think</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>The more that you read</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>the more things you will know</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>The more that you learn</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>the more places you’ll go</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>the mat has a cat with a hat</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              text\n",
       "0              the cat with the hat sat in the mat\n",
       "1                    you have brains in your head \n",
       "2                      you have feet in your shoes\n",
       "3  You can steer yourself any direction you choose\n",
       "4               Think and wonder, wonder and think\n",
       "5                           The more that you read\n",
       "6                    the more things you will know\n",
       "7                          The more that you learn\n",
       "8                        the more places you’ll go\n",
       "9                     the mat has a cat with a hat"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "#dr seuss quotes!!! some examples of text\n",
    "\n",
    "text = [ \n",
    "    \n",
    "        'the cat with the hat sat in the mat',\n",
    "        'you have brains in your head ',' you have feet in your shoes',\n",
    "        'You can steer yourself any direction you choose' ,\n",
    "        'Think and wonder, wonder and think', 'The more that you read',' the more things you will know',\n",
    "        'The more that you learn' ,'the more places you’ll go',\n",
    "        'the mat has a cat with a hat']\n",
    "\n",
    "#from list into a pandas dataframe\n",
    "txt = pd.DataFrame(text,columns=['text'])\n",
    "\n",
    "txt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Counts/Occurrences\n",
    "\n",
    "The model is simple in that it throws away all of the order information in the words and focuses on the occurrence of words in a document.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,\n",
       "        0, 0, 0, 3, 0, 0, 0, 1, 0, 0, 0, 0],\n",
       "       [0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "        0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0],\n",
       "       [0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "        1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0],\n",
       "       [0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "        0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1],\n",
       "       [2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "        0, 0, 0, 0, 0, 2, 0, 0, 2, 0, 0, 0],\n",
       "       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0,\n",
       "        0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0],\n",
       "       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0,\n",
       "        0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 0],\n",
       "       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0,\n",
       "        0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0],\n",
       "       [0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0,\n",
       "        0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0],\n",
       "       [0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,\n",
       "        0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0]], dtype=int64)"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "\n",
    "count_vect = CountVectorizer()\n",
    "X_transformed = count_vect.fit_transform(txt.text.values)\n",
    "\n",
    "\n",
    "# by using .A we show the sparse matrix as an numpy array\n",
    "X_transformed.A\n",
    "\n",
    "\n",
    "#This is the format machine learning algorithms understand from text data!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### But this is goobledygook ! what does it REALLY MEAN?\n",
    "\n",
    "in order to understand what is happening we  create a dataframe from the matrix above to elucidate the Countvectorizer process. We append a new column with the feature names with .get_feature_names() to make it even more clear."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>term</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>and</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>any</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>brains</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>can</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>cat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>choose</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>direction</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>feet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>go</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>has</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>hat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>have</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>head</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>in</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>know</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>learn</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>ll</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>mat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>more</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>places</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>read</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>sat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>shoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>steer</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>that</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>the</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>things</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>think</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>will</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>with</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>wonder</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>you</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>your</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>yourself</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    0  1  2  3  4  5  6  7  8  9       term\n",
       "0   0  0  0  0  2  0  0  0  0  0        and\n",
       "1   0  0  0  1  0  0  0  0  0  0        any\n",
       "2   0  1  0  0  0  0  0  0  0  0     brains\n",
       "3   0  0  0  1  0  0  0  0  0  0        can\n",
       "4   1  0  0  0  0  0  0  0  0  1        cat\n",
       "5   0  0  0  1  0  0  0  0  0  0     choose\n",
       "6   0  0  0  1  0  0  0  0  0  0  direction\n",
       "7   0  0  1  0  0  0  0  0  0  0       feet\n",
       "8   0  0  0  0  0  0  0  0  1  0         go\n",
       "9   0  0  0  0  0  0  0  0  0  1        has\n",
       "10  1  0  0  0  0  0  0  0  0  1        hat\n",
       "11  0  1  1  0  0  0  0  0  0  0       have\n",
       "12  0  1  0  0  0  0  0  0  0  0       head\n",
       "13  1  1  1  0  0  0  0  0  0  0         in\n",
       "14  0  0  0  0  0  0  1  0  0  0       know\n",
       "15  0  0  0  0  0  0  0  1  0  0      learn\n",
       "16  0  0  0  0  0  0  0  0  1  0         ll\n",
       "17  1  0  0  0  0  0  0  0  0  1        mat\n",
       "18  0  0  0  0  0  1  1  1  1  0       more\n",
       "19  0  0  0  0  0  0  0  0  1  0     places\n",
       "20  0  0  0  0  0  1  0  0  0  0       read\n",
       "21  1  0  0  0  0  0  0  0  0  0        sat\n",
       "22  0  0  1  0  0  0  0  0  0  0      shoes\n",
       "23  0  0  0  1  0  0  0  0  0  0      steer\n",
       "24  0  0  0  0  0  1  0  1  0  0       that\n",
       "25  3  0  0  0  0  1  1  1  1  1        the\n",
       "26  0  0  0  0  0  0  1  0  0  0     things\n",
       "27  0  0  0  0  2  0  0  0  0  0      think\n",
       "28  0  0  0  0  0  0  1  0  0  0       will\n",
       "29  1  0  0  0  0  0  0  0  0  1       with\n",
       "30  0  0  0  0  2  0  0  0  0  0     wonder\n",
       "31  0  1  1  2  0  1  1  1  1  0        you\n",
       "32  0  1  1  0  0  0  0  0  0  0       your\n",
       "33  0  0  0  1  0  0  0  0  0  0   yourself"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vectdf =pd.DataFrame(X_transformed.A.T)\n",
    "vectdf['term']=count_vect.get_feature_names()\n",
    "\n",
    "#rows are words. each column represents a document. This is the DTM *document term matrix*\n",
    "vectdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'and': 0,\n",
       " 'any': 1,\n",
       " 'brains': 2,\n",
       " 'can': 3,\n",
       " 'cat': 4,\n",
       " 'choose': 5,\n",
       " 'direction': 6,\n",
       " 'feet': 7,\n",
       " 'go': 8,\n",
       " 'has': 9,\n",
       " 'hat': 10,\n",
       " 'have': 11,\n",
       " 'head': 12,\n",
       " 'in': 13,\n",
       " 'know': 14,\n",
       " 'learn': 15,\n",
       " 'll': 16,\n",
       " 'mat': 17,\n",
       " 'more': 18,\n",
       " 'places': 19,\n",
       " 'read': 20,\n",
       " 'sat': 21,\n",
       " 'shoes': 22,\n",
       " 'steer': 23,\n",
       " 'that': 24,\n",
       " 'the': 25,\n",
       " 'things': 26,\n",
       " 'think': 27,\n",
       " 'will': 28,\n",
       " 'with': 29,\n",
       " 'wonder': 30,\n",
       " 'you': 31,\n",
       " 'your': 32,\n",
       " 'yourself': 33}"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Show the words and the its numerical index\n",
    "count_vect.vocabulary_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### From occurrences to frequencies\n",
    "\n",
    "Occurrence count is a good start but there is an issue: longer documents will have higher average count \n",
    "values than shorter documents, even though they might talk about the same topics.\n",
    "\n",
    "To avoid these potential discrepancies it suffices to divide the number of occurrences of each word in a document by the total number of words in the document: these new features are called tf for Term Frequencies.\n",
    "\n",
    "Another refinement on top of tf is to downscale weights for words that occur in many documents in the corpus and are therefore less informative than those that occur only in a smaller portion of the corpus.\n",
    "\n",
    "     This downscaling is called tf–idf for “Term Frequency times Inverse Document Frequency”."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Term Frequency (TF) \n",
    "\n",
    "> TF for a term “t” is defined as the count of a term “t” in a document “D”\n",
    "\n",
    "#### Inverse Document Frequency (IDF) \n",
    "> IDF for a term is defined as logarithm of ratio of total documents available in the corpus and number of documents  containing the term T."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>term</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.57735</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>and</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>any</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.485869</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>brains</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>can</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.320449</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.415853</td>\n",
       "      <td>cat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>choose</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>direction</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.485869</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>feet</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>go</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.489186</td>\n",
       "      <td>has</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.320449</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.415853</td>\n",
       "      <td>hat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.413033</td>\n",
       "      <td>0.413033</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>have</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.485869</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>head</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.280355</td>\n",
       "      <td>0.361355</td>\n",
       "      <td>0.361355</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>in</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>know</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.610201</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>learn</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>ll</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.320449</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.415853</td>\n",
       "      <td>mat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.403482</td>\n",
       "      <td>0.332153</td>\n",
       "      <td>0.403482</td>\n",
       "      <td>0.332153</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>more</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>places</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.610201</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>read</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.376958</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>sat</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.485869</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>shoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>steer</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.518727</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.518727</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>that</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.607086</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.327573</td>\n",
       "      <td>0.269663</td>\n",
       "      <td>0.327573</td>\n",
       "      <td>0.269663</td>\n",
       "      <td>0.262609</td>\n",
       "      <td>the</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>things</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.57735</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>think</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.502327</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>will</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.320449</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.415853</td>\n",
       "      <td>with</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.57735</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>wonder</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.236841</td>\n",
       "      <td>0.236841</td>\n",
       "      <td>0.369795</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.297448</td>\n",
       "      <td>0.244864</td>\n",
       "      <td>0.297448</td>\n",
       "      <td>0.244864</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>you</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.413033</td>\n",
       "      <td>0.413033</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>your</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.379309</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>yourself</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           0         1         2         3        4         5         6  \\\n",
       "0   0.000000  0.000000  0.000000  0.000000  0.57735  0.000000  0.000000   \n",
       "1   0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "2   0.000000  0.485869  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "3   0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "4   0.320449  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "5   0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "6   0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "7   0.000000  0.000000  0.485869  0.000000  0.00000  0.000000  0.000000   \n",
       "8   0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "9   0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "10  0.320449  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "11  0.000000  0.413033  0.413033  0.000000  0.00000  0.000000  0.000000   \n",
       "12  0.000000  0.485869  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "13  0.280355  0.361355  0.361355  0.000000  0.00000  0.000000  0.000000   \n",
       "14  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.502327   \n",
       "15  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "16  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "17  0.320449  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "18  0.000000  0.000000  0.000000  0.000000  0.00000  0.403482  0.332153   \n",
       "19  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "20  0.000000  0.000000  0.000000  0.000000  0.00000  0.610201  0.000000   \n",
       "21  0.376958  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "22  0.000000  0.000000  0.485869  0.000000  0.00000  0.000000  0.000000   \n",
       "23  0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "24  0.000000  0.000000  0.000000  0.000000  0.00000  0.518727  0.000000   \n",
       "25  0.607086  0.000000  0.000000  0.000000  0.00000  0.327573  0.269663   \n",
       "26  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.502327   \n",
       "27  0.000000  0.000000  0.000000  0.000000  0.57735  0.000000  0.000000   \n",
       "28  0.000000  0.000000  0.000000  0.000000  0.00000  0.000000  0.502327   \n",
       "29  0.320449  0.000000  0.000000  0.000000  0.00000  0.000000  0.000000   \n",
       "30  0.000000  0.000000  0.000000  0.000000  0.57735  0.000000  0.000000   \n",
       "31  0.000000  0.236841  0.236841  0.369795  0.00000  0.297448  0.244864   \n",
       "32  0.000000  0.413033  0.413033  0.000000  0.00000  0.000000  0.000000   \n",
       "33  0.000000  0.000000  0.000000  0.379309  0.00000  0.000000  0.000000   \n",
       "\n",
       "           7         8         9       term  \n",
       "0   0.000000  0.000000  0.000000        and  \n",
       "1   0.000000  0.000000  0.000000        any  \n",
       "2   0.000000  0.000000  0.000000     brains  \n",
       "3   0.000000  0.000000  0.000000        can  \n",
       "4   0.000000  0.000000  0.415853        cat  \n",
       "5   0.000000  0.000000  0.000000     choose  \n",
       "6   0.000000  0.000000  0.000000  direction  \n",
       "7   0.000000  0.000000  0.000000       feet  \n",
       "8   0.000000  0.502327  0.000000         go  \n",
       "9   0.000000  0.000000  0.489186        has  \n",
       "10  0.000000  0.000000  0.415853        hat  \n",
       "11  0.000000  0.000000  0.000000       have  \n",
       "12  0.000000  0.000000  0.000000       head  \n",
       "13  0.000000  0.000000  0.000000         in  \n",
       "14  0.000000  0.000000  0.000000       know  \n",
       "15  0.610201  0.000000  0.000000      learn  \n",
       "16  0.000000  0.502327  0.000000         ll  \n",
       "17  0.000000  0.000000  0.415853        mat  \n",
       "18  0.403482  0.332153  0.000000       more  \n",
       "19  0.000000  0.502327  0.000000     places  \n",
       "20  0.000000  0.000000  0.000000       read  \n",
       "21  0.000000  0.000000  0.000000        sat  \n",
       "22  0.000000  0.000000  0.000000      shoes  \n",
       "23  0.000000  0.000000  0.000000      steer  \n",
       "24  0.518727  0.000000  0.000000       that  \n",
       "25  0.327573  0.269663  0.262609        the  \n",
       "26  0.000000  0.000000  0.000000     things  \n",
       "27  0.000000  0.000000  0.000000      think  \n",
       "28  0.000000  0.000000  0.000000       will  \n",
       "29  0.000000  0.000000  0.415853       with  \n",
       "30  0.000000  0.000000  0.000000     wonder  \n",
       "31  0.297448  0.244864  0.000000        you  \n",
       "32  0.000000  0.000000  0.000000       your  \n",
       "33  0.000000  0.000000  0.000000   yourself  "
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_vect = TfidfVectorizer()\n",
    "X_tf_transformed = tfidf_vect.fit_transform(txt.text.values)\n",
    "\n",
    "\n",
    "# this is what a machine learning algorithm needs.\n",
    "\n",
    "\n",
    "#but for better understanding... lets see what is the DTM. This time its \n",
    "#the DTM holds downscaled frequencies as values and not counts.\n",
    "\n",
    "tfidf_vectdf =pd.DataFrame(X_tf_transformed.A.T)\n",
    "tfidf_vectdf['term']=tfidf_vect.get_feature_names()\n",
    "tfidf_vectdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  (0, 7)\t0.5844829010200651\n",
      "  (0, 2)\t0.5844829010200651\n",
      "  (0, 4)\t0.444514311537431\n",
      "  (0, 1)\t0.34520501686496574\n",
      "  (1, 1)\t0.3853716274664007\n",
      "  (1, 0)\t0.652490884512534\n",
      "  (1, 3)\t0.652490884512534\n",
      "  (2, 4)\t0.444514311537431\n",
      "  (2, 1)\t0.34520501686496574\n",
      "  (2, 6)\t0.5844829010200651\n",
      "  (2, 5)\t0.5844829010200651\n"
     ]
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "obj = TfidfVectorizer()\n",
    "obj = TfidfVectorizer()\n",
    "\n",
    "corpus = ['This is sample document.', 'another random document.', 'third sample document text']\n",
    "\n",
    "X = obj.fit_transform(corpus)\n",
    "\n",
    "print(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "# Advanced NLP tasks\n",
    "\n",
    "---\n",
    "\n",
    "### Text Classification\n",
    "\n",
    "\n",
    "Text classification, in common words is defined as a technique to systematically classify a text object (document or sentence) in one of the fixed category. It is really helpful when the amount of data is too large, especially for organizing, information filtering, and storage purposes.\n",
    "\n",
    "A typical natural language classifier consists of two parts: *Training* and *Prediction* \n",
    "(as shown in image below) \n",
    "\n",
    "Firstly the text input is processes and features are created. The machine learning models then learn these features and is used for predicting against the new text.\n",
    "\n",
    "Examples of text classification applications are: Email Spam Identification, topic classification of news, sentiment classification and organization of web pages by search engines.\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### you will need to install textblob\n",
    "\n",
    "```\n",
    "pip install -U textblob\n",
    "```\n",
    "\n",
    "#### and download the corpora\n",
    "\n",
    "```\n",
    "python -m textblob.download_corpora\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'textblob'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-60-ab135d45aff1>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mtextblob\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclassifiers\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mNaiveBayesClassifier\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mNBC\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtextblob\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mTextBlob\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m training_corpus = [ ('I am exhausted of this work.', 'Class_B'),\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'textblob'"
     ]
    }
   ],
   "source": [
    "from textblob.classifiers import NaiveBayesClassifier as NBC\n",
    "from textblob import TextBlob\n",
    "\n",
    "\n",
    "training_corpus = [ ('I am exhausted of this work.', 'Class_B'),\n",
    "                   (\"I can't cooperate with this\", 'Class_B'),\n",
    "                   ('He is my badest enemy!', 'Class_B'),\n",
    "                   ('My management is poor.', 'Class_B'),\n",
    "                   ('I love this burger.', 'Class_A'),\n",
    "                   ('This is an brilliant place!', 'Class_A'),\n",
    "                   ('I feel very good about these dates.', 'Class_A'),\n",
    "                   ('This is my best work.', 'Class_A'),\n",
    "                   (\"What an awesome view\", 'Class_A'),\n",
    "                   ('I do not like this dish', 'Class_B')]\n",
    "\n",
    "test_corpus = [(\"I am not feeling well today.\", 'Class_B'), \n",
    "                (\"I feel brilliant!\", 'Class_A'), \n",
    "                ('Gary is a friend of mine.', 'Class_A'), \n",
    "                (\"I can't believe I'm doing this.\", 'Class_B'), \n",
    "                ('The date was good.', 'Class_A'), ('I do not enjoy my job', 'Class_B')]\n",
    "\n",
    "model = NBC(training_corpus) \n",
    "\n",
    "print(model.classify(\"Their codes are amazing.\"))\n",
    "print(model.classify(\"I don't like their computer.\"))\n",
    "\n",
    "print (model.classify(test_corpus[4][0]))\n",
    "print(model.accuracy(test_corpus))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'training_corpus' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-61-ced30d032adf>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0mtrain_labels\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0mrow\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mtraining_corpus\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m     \u001b[0mtrain_data\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrow\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m     \u001b[0mtrain_labels\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrow\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'training_corpus' is not defined"
     ]
    }
   ],
   "source": [
    "import sklearn \n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn import svm\n",
    "\n",
    "# preparing data for SVM model (using the same training_corpus, test_corpus from naive bayes example)\n",
    "train_data = []\n",
    "train_labels = []\n",
    "\n",
    "for row in training_corpus:\n",
    "    train_data.append(row[0])\n",
    "    train_labels.append(row[1])\n",
    "\n",
    "test_data = [] \n",
    "test_labels = []\n",
    "\n",
    "for row in test_corpus:\n",
    "    test_data.append(row[0]) \n",
    "    test_labels.append(row[1])\n",
    "\n",
    "# Create feature vectors \n",
    "vectorizer = TfidfVectorizer(min_df=4, max_df=0.9)\n",
    "\n",
    "# Train the feature vectors\n",
    "train_vectors = vectorizer.fit_transform(train_data)\n",
    "\n",
    "# Apply model on test data \n",
    "test_vectors = vectorizer.transform(test_data)\n",
    "\n",
    "# Perform classification with SVM, kernel=linear \n",
    "model = svm.SVC(kernel='linear') \n",
    "model.fit(train_vectors, train_labels) \n",
    "prediction = model.predict(test_vectors)\n",
    "\n",
    "\n",
    "print (classification_report(test_labels, prediction))\n",
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Text Matching / Similarity\n",
    "\n",
    "#### A. Levenshtein Distance \n",
    "The Levenshtein distance between two strings is defined as \n",
    "\n",
    "> the minimum number of edits needed to transform one string into the other,\n",
    "> with the allowable edit operations being insertion, deletion, or substitution of a single character. \n",
    "\n",
    "Following is the implementation for efficient memory computations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "9\n"
     ]
    }
   ],
   "source": [
    "def levenshtein(s1,s2): \n",
    "    if len(s1) > len(s2):\n",
    "        s1,s2 = s2,s1 \n",
    "    distances = range(len(s1) + 1) \n",
    "    for index2,char2 in enumerate(s2):\n",
    "        newDistances = [index2+1]\n",
    "        for index1,char1 in enumerate(s1):\n",
    "            if char1 == char2:\n",
    "                newDistances.append(distances[index1]) \n",
    "            else:\n",
    "                 newDistances.append(1 + min((distances[index1], distances[index1+1], newDistances[-1]))) \n",
    "        distances = newDistances \n",
    "    return distances[-1]\n",
    "\n",
    "print(levenshtein(\"analyze\",\"analyse\"))\n",
    "\n",
    "print(levenshtein(\"analyze\",\"information\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "### Phonetic Matching \n",
    "\n",
    "A Phonetic matching algorithm takes a keyword as input (person’s name, location name etc) and produces a character string that identifies a set of words that are (roughly) phonetically similar. It is very useful for searching large text corpuses, correcting spelling errors and matching relevant names. \n",
    "\n",
    "**Soundex** and **Metaphone**\n",
    "are two main phonetic algorithms used for this purpose. \n",
    "\n",
    "Python’s module Fuzzy is used to compute soundex strings for different words, for example –"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'fuzzywuzzy'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-64-d0dcbfb75236>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mfuzzywuzzy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mfuzz\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mfuzzywuzzy\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mprocess\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mfuzz\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mratio\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"ACME Factory alpha\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"ACME Factory Inc.\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mfuzz\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpartial_ratio\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"ACME Factory alpha\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"ACME Factory Inc.\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'fuzzywuzzy'"
     ]
    }
   ],
   "source": [
    "from fuzzywuzzy import fuzz\n",
    "from fuzzywuzzy import process\n",
    "\n",
    "fuzz.ratio(\"ACME Factory alpha\", \"ACME Factory Inc.\")\n",
    "fuzz.partial_ratio(\"ACME Factory alpha\", \"ACME Factory Inc.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'fuzzy'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-66-9163cb0da355>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mfuzzy\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0msoundex\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfuzzy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mSoundex\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m4\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msoundex\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'ankita'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'fuzzy'"
     ]
    }
   ],
   "source": [
    "import fuzzy\n",
    "\n",
    "soundex = fuzzy.Soundex(4)\n",
    "\n",
    "print(soundex('ankita'))\n",
    "print(soundex('aunkit'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Text Classification some basics\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAGsCAYAAAAsSYsFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzt3XmcHWWd7/HPF0JYBGQJo5EtQYOKGCAJi4BsLiMOyBVRRLiyXdGLMo77Mo4g6jiOjuM6YNwQuYCyqCxRNlmGkbUhBhDRyCIRFBBBMGyB7/2j6sCh7XQ13V2nuk99369Xv/qcqnPO80te1f3teuqp55FtIiIihrNC0wVERMTEl7CIiIhKCYuIiKiUsIiIiEoJi4iIqJSwiIiISgmLiAlI0lGSTmi6joiOhEVERFRKWERERKWERcQzJGmOpGslPSDpFEnfl/Spct/bJC2WdK+kMyQ9r+t920u6StL95fftu/bNlHRx+ZnnAdMa+KdFLFfCIuIZkDQV+CFwHLAOcBLw+nLfbsBngDcB04HbgJPLfesAZwNfBtYFvgCcLWnd8qNPBAYoQuKTwIE9+QdFjJAyN1TEyEnaiSIgNnD5wyPpUuAiioD4k+0PlttXB/4MzAJeDhxhe5uuz7oM+DrwM+Bm4Nm2/1ruOxF4wvYBPfqnRQwrZxYRz8zzgN/76X9l3d6177bORtsPAn8C1h+8r3Rb174/d4Kia1/EhJGwiHhm7gTWl6SubRuW3+8ANu5slPQsii6n3w/eV9qo3HcnsHb5+u59ERNGwiLimbkMeBx4l6QpkvYCOl1LJwIHS9pS0srAvwJX2L4VWABsKukt5fv2BTYDzrJ9G3A18AlJUyXtCOzZ439XxLASFhHPgO1Hgb2BQ4H7gAOAs4BHbF8A/AtwGsXZwvOBN5fv+xOwB/A+iq6pDwJ72L6n/Oi3ANsC9wJHAsf36J8UMSK5wB0xRpKuAI61/Z2ma4moS84sIp4hSTtLem7ZnXQgMBv4adN1RdRpStMFRExCLwR+AKwO/BbYx/adzZYUUa90Q0VERKV0Q0VERKWERUREVOqbaxbTpk3zjBkzRv3+hx56iFVXXXX8CorokuMr6jSW42tgYOAe2+tVva5vwmLGjBlcffXVo37/wMAAc+fOHceKIp6S4yvqNJbjS9KIppZJN1RERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESl2sJC0rcl3SXp+uXsl6QvS1osaZGkOYP2rynp95K+WleNERExMnWeWRwHvGaY/bsDs8qvw4BjBu3/JHBxLZVFRMQzUltY2L4EuHeYl+wFHO/C5cBakqYDSJoLPAc4t676IiJi5Jq8ZrE+cHvX8yXA+pJWAP4D+EAjVUVExN+Y0mDbGmKbgcOBBbZvl4Z6SdcHSIdRdGExffp0BgYGRl3M0qVLx/T+iOHk+Io69eL4ajIslgAbdj3fALgDeBnwckmHA6sDUyU9aPvDgz/A9nxgPsC8efM8d+7cURczMDDAWN4fMZwcX1GnXhxfTYbFGcC7JJ0MbAvcb/tOYP/OCyQdBMwbKigiIqJ3agsLSScBuwDTJC0BjgRWArB9LLAAeC2wGFgKHFxXLRERMTa1hYXt/Sr2G3hnxWuOoxiCGxERDcod3BERUSlhERERlRIWERFRKWERERGVEhYREVEpYREREZUSFhERUSlhERERlRIWERFRKWERERGVEhYREVEpYREREZUSFhERUSlhERERlRIWERFRKWERERGVEhYREVEpYREREZUSFhERUSlhERERlRIWERFRKWERERGVEhYREVEpYREREZUSFhERUSlhERERlRIWERFRKWERERGVEhYREVEpYREREZVqCwtJ35Z0l6Trl7Nfkr4sabGkRZLmlNu3lHSZpBvK7fvWVWNERIxMnWcWxwGvGWb/7sCs8usw4Jhy+1LgrbZfUr7/i5LWqrHOiIioMKWuD7Z9iaQZw7xkL+B42wYul7SWpOm2f931GXdIugtYD7ivrlojImJ4TV6zWB+4vev5knLbkyRtA0wFftvDuiIiYpDazixGQENs85M7penA94ADbT8x5AdIh1F0YTF9+nQGBgZGXczSpUvH9P6I4eT4ijr14vhqMiyWABt2Pd8AuANA0prA2cDHbF++vA+wPR+YDzBv3jzPnTt31MUMDAwwlvdHDCfHV9SpF8dXk91QZwBvLUdFbQfcb/tOSVOBH1JczzilwfoiIqJU25mFpJOAXYBpkpYARwIrAdg+FlgAvBZYTDEC6uDyrW8CdgLWlXRQue0g2wvrqjUiIoZX52io/Sr2G3jnENtPAE6oq66IiHjmcgd3RERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUesZhIWltSbPrKCYiIiamEYWFpIskrSlpHeAXwHckfaHe0iIiYqIY6ZnFs23/Bdgb+I7tucAr6ysrIiImkpGGxRRJ04E3AWfVWE9ERExAIw2LTwDnAIttXyVpE+A39ZUVERETyZQRvu5O209e1LZ9c65ZRES0x0jPLL4ywm0REdGHhj2zkPQyYHtgPUnv7dq1JrBinYVFRMTEUdUNNRVYvXzdGl3b/wLsU1dRERExsQwbFrYvBi6WdJzt23pUU0RETDAjvcC9sqT5wIzu99jerY6iIiJiYhlpWJwCHAt8E3i8vnIiImIiGuloqGW2j7F9pe2Bztdwb5D0bUl3Sbp+Ofsl6cuSFktaJGlO174DJf2m/DrwGfx7IiKiBiMNizMlHS5puqR1Ol8V7zkOeM0w+3cHZpVfhwHHAJSfeySwLbANcKSktUdYZ0RE1GCk3VCdv+4/0LXNwCbLe4PtSyTNGOYz9wKOt23gcklrlVOK7AKcZ/teAEnnUYTOSSOsNSIixtmIwsL2zBraXh+4vev5knLb8rbXZsaHzy4enHJ2nc1E2+X4ihrdOrfezx9RWEh661DbbR8/hrY11EcOs32oug6j6MJi+vTpDAwMexklIqJv1f37b6TdUFt3PV4FeAVwDTCWsFgCbNj1fAPgjnL7LoO2XzTUB9ieD8wHmDdvnufOHV203jq3+I8e7fsjquT4ijr14vgaaTfUEd3PJT0b+N4Y2z4DeJekkykuZt9v+05J5wD/2nVR+9XAR8bYVkREjMFIzywGW0oximm5JJ1EcYYwTdISihFOKwHYPhZYALwWWFx+3sHlvnslfRK4qvyoozsXuyMiohkjvWZxJk9dN1gReDHwg+HeY3u/iv0G3rmcfd8Gvj2S2iIion4jPbP4fNfjZcBttpfUUE9ERExAI7opr5xQ8FcUM8+uDTxaZ1ERETGxjCgsJL0JuBJ4I8U63FdIyhTlEREtMdJuqH8GtrZ9F4Ck9YDzgVPrKiwiIiaOkc4NtUInKEp/egbvjYiISW6kZxY/Le9/6MzPtC/F0NeIiGiBqjW4XwA8x/YHJO0N7EgxHcdlwP/rQX0RETEBVHUlfRF4AMD26bbfa/s9FGcVX6y7uIiImBiqwmKG7UWDN9q+mmKJ1YiIaIGqsFhlmH2rjmchERExcVWFxVWS3jZ4o6RDgcwHHhHRElWjof4J+KGk/XkqHOYBU4HX11lYRERMHMOGhe0/AttL2hXYvNx8tu2f1V5ZRERMGCNdz+JC4MKaa4mIiAkqd2FHRESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVKo1LCS9RtJNkhZL+vAQ+zeWdIGkRZIukrRB175/l3SDpBslfVmS6qw1IiKWr7awkLQi8DVgd2AzYD9Jmw162eeB423PBo4GPlO+d3tgB2A2xXKuWwM711VrREQMr84zi22AxbZvtv0ocDKw16DXbAZcUD6+sGu/gVWAqcDKwErAH2usNSIihlFnWKwP3N71fEm5rdsvgDeUj18PrCFpXduXUYTHneXXObZvrLHWiIgYxpQaP3uoawwe9Pz9wFclHQRcAvweWCbpBcCLgc41jPMk7WT7kqc1IB0GHAYwffp0BgYGRl3s0qVLx/T+iOHk+Io69eL4qjMslgAbdj3fALij+wW27wD2BpC0OvAG2/eXIXC57QfLfT8BtqMIlO73zwfmA8ybN89z584ddbEDAwOM5f0Rw8nxFXXqxfFVZzfUVcAsSTMlTQXeDJzR/QJJ0yR1avgI8O3y8e+AnSVNkbQSxcXtdENFRDSktrCwvQx4F3AOxS/6H9i+QdLRkl5XvmwX4CZJvwaeA3y63H4q8FvgOorrGr+wfWZdtUZExPDq7IbC9gJgwaBtH+96fCpFMAx+3+PA2+usLSIiRi53cEdERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKVaw0LSayTdJGmxpA8PsX9jSRdIWiTpIkkbdO3bSNK5km6U9EtJM+qsNSIilq+2sJC0IvA1YHdgM2A/SZsNetnngeNtzwaOBj7Tte944HO2XwxsA9xVV60RETG8Os8stgEW277Z9qPAycBeg16zGXBB+fjCzv4yVKbYPg/A9oO2l9ZYa0REDKPOsFgfuL3r+ZJyW7dfAG8oH78eWEPSusCmwH2STpd0raTPlWcqERHRgCk1fraG2OZBz98PfFXSQcAlwO+BZWVdLwe2An4HfB84CPjW0xqQDgMOA5g+fToDAwOjLnbp0qVjen/EcHJ8RZ16cXzVGRZLgA27nm8A3NH9Att3AHsDSFodeIPt+yUtAa61fXO570fAdgwKC9vzgfkA8+bN89y5c0dd7MDAAGN5f8RwcnxFnXpxfNXZDXUVMEvSTElTgTcDZ3S/QNI0SZ0aPgJ8u+u9a0tar3y+G/DLGmuNiIhh1BYWtpcB7wLOAW4EfmD7BklHS3pd+bJdgJsk/Rp4DvDp8r2PU3RRXSDpOoourW/UVWtERAyvzm4obC8AFgza9vGux6cCpy7nvecBs+usLyIiRiZ3cEdERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFRKWERERKWERUREVEpYREREpYRFRERUSlhERESlhEVERFSS7aZrGBeS7gZuG8NHTAPuGadyIgbL8RV1GsvxtbHt9ape1DdhMVaSrrY9r+k6oj/l+Io69eL4SjdURERUSlhERESlhMVT5jddQPS1HF9Rp9qPr1yziIiISjmziIiISgmLiIiolLCIiIhKU5ouoAmSVgH2AF4OPA94CLgeONv2DU3WFv1F0rOAh20/3nQt0T8kzeNvf3+db/ve2tps2wVuSUcBewIXAQPAXcAqwKbAruXj99le1FCJMYlJWgF4M7A/sDXwCLAycDewAJhv+zfNVRiTmaSDgH8EbuFvf3/tQBEa/2L7d+PddhvPLK6yfdRy9n1B0t8BG/WwnugvFwLnAx8Brrf9BICkdSj+GPk3ST+0fUKDNcbk9SxgB9sPDbVT0pbALGDcw6J1ZxYRdZK0ku3HxvqaiImmdWEh6TuAgfttv6fpeiIiRkrSx8uHD9r+Qi/bbmM31HHl90ebLCL6k6RbKP4Yudv2tk3XE32nM7P2kN1QdWrdmUU3SasCG9m+qelaIiImsjaeWQAgaU/g88BUYGZ5Yeho269rtrKYzCTNGW6/7Wt6VUv0L0mbAh8ANqbr97jt3Wprs61nFpIGgN2Ai2xvVW5bZHt2s5XFZCbpwvLhKsA84BeAgNnAFbZ3bKq26B+SfgEcSzF89sl7eGwP1NVma88sgGW275fUdB3RR2zvCiDpZOAw29eVzzcH3t9kbdFXltk+ppcNtjksrpf0FmBFSbMobnT5ecM1Rf94UScoAGxfX3Z1Roxaeb8OwJmSDgd+SHHjJwC5g7sGklYD/hl4NUU3wTnAJ20/3Ghh0RcknQT8FTiBYnTUAcDqtvdrtLCY1LpG2w3VJWLbm9TWdlvDIqJO5fxj/xfYqdx0CXBM/hiJyap1YSHpTIpkHlJGQ8V4ydDsqIukvYfYfD9wne27ammzhWGx83D7bV/cq1qif0l6HfA5YKrtDM2OcSXpbOBlFHORAewCXE4xoeDRtr833m227gJ3dxhImgq8iOJM4ybbuas7xsuRwDYUsxtje6GkGQ3WE/3lCeDFtv8IIOk5wDHAthRdngmL8SLpHyjGKf+W4mLRTElvt/2TZiuLPpGh2VGnGZ2gKN0FbGr7Xkm1TFLZ2rAA/gPY1fZiAEnPB84GEhYxHjI0O+r035LOAk4pn78BuKRcbOu+Ohps3TWLDkmX2N6p67mAi7u3RYxWhmZHncrfV2+gWPBIwKXAaa7xF3rrwqJrFMGrKOZV+QHFNYs3Uly3eF9TtUVETFRt7Ibas+vxH4HO6Ki7gbV7X070kwzNjl6Q9ABPHWdTgZWAv9pes642WxcWtg9uuoboa58vv+8NPJfiDm6A/YBbmygo+o/tNbqfS/pfFKPvatO6bqiO8g7bQ4GXUMwQCoDtQxorKvrG4Gtiy9sWMV4kXW57u7o+v3VnFl2+B/wK+HvgaGB/4MZGK4p+sp6kTWzfDCBpJrBewzVFnxh0B/cKFNPh1/qXf5vD4gW23yhpL9vflXQixYiViPHwHuAiSTeXz2cAb2+unOgz3ddel1F0ce5VZ4NtDovOjSv3lWsN/IHiBzpizGz/tLy/4kXlpl/ZfmS490SMVBPXXlfodYMTyHxJawMfA84Afgl8ttmSos/MpbgmtgWwr6S3NlxP9AlJG0j6oaS7JP1R0mmSNqi1zRZf4J5p+5aqbRGjIel7wPOBhTy17KVt/2NzVUW/kHQecCJPzQF1ALC/7VfV1maLw+Ia23MGbRuwPbepmqJ/SLoR2KzOO2qjvSQttL1l1bbx1LprFpJeRNE18OxBIwrWpGsIbcQYXU9xn8WdTRcSfekeSQcAJ5XP9wP+VGeDrQsL4IXAHsBaPH1EwQPA2xqpKPrRNOCXkq7k6Wsk5w7uGA+HAF8F/pNiyOzPy221aXM31MtsX9Z0HdGflrfIVhbXismqdWEh6YO2/13SVxjiJpZcgIzxImljYJbt88tZaFe0/UDTdcXkJ2k9ip6QGXT1ENU5A0Ubu6E6d2lf3WgV0dckvQ04DFiHYlTU+hSLbb2iybqib/wY+G/gfJ4abVer1p1ZRPSCpIUUE7tdYXurctt1tl/abGXRD+oe+TSUNp5ZACDpQobuhtqtgXKi/zxi+9HOsqqSplDz3D3RKmdJeq3tBb1qsLVhAby/6/EqFKtOLWuolug/F0v6KLCqpFcBhwNnNlxT9I93Ax+V9CjwKMVqea5zPYt0Q3WRdLHtIUexRDwTklagmAL/yWVVbX+j2aoiRq+1ZxaS1ul6ugLFPD7Pbaic6D9H2P4S8GRASHp3uS1iTMo1uPcHZtr+pKQNgem2r6ytzbaeWUi6haIPWRTdT7cAR9u+tNHCoi8sZzqZazsXuyPGQtIxwBPAbrZfXE6Keq7tretqs7VnFrZnNl1D9B9J+wFvAWZKOqNr1xrUPB1DtMq2tudIuhbA9p8lTa2zwdaGhaQ3Aj+1/YCkjwFzgE/Zvqbh0mJy+znFfFDTgP/o2v4AsKiRiqIfPSZpRcoRduVNek/U2WCbu6EW2Z4taUfgM8DngY/a3rbh0iIihiVpf2Bfij9yvwvsA3zM9im1tdnisLjW9laSPgNcZ/vE9CnHeClnNP4s8HcU18VqH9oY7VLOoP0KimPrAts3VrxlbO21OCzOAn4PvJJiJNRDwJW2t2i0sOgLkhYDe9b9AxzRK20Oi9WA11CcVfxG0nTgpbbPbbi06AOS/sf2Dk3XEf2laxTn3b3uMm9tWETUSdKXKO7b+RFPX8/i9MaKihiD1o2GajKZo1XWBJZS3MHdYSBhEZNSziwiIqJS684sIuqUxbWiXyUsIsZXFteKvpRuqIiISU5S54+Ur9n+ah1t5MwiYhxJ+g5F99P9tt/TdD3RDuVkgusC29XVRs4sSr1I5uh/kjrroTxq+7JGi4kYRwmLLp1ktn1207VERAyWm/IiImJCa901i9yUFxGT1aAVPv+G7XtraztnFhERk8OgFT43Av5cPl4L+F2di7qtUNcHTwaSdpR0cPl4PUlZPS9qIelfJX2ovC4WMSq2Z9reBDiHYlbjabbXBfag5qlkWhsWko4EPgR8pNy0EnBCcxVFn7uSYq33/2y6kOgLW9te0Hli+yfAzsO8fsxad82iy+uBrYBrAGzfIWmNZkuKfmX7R03XEH3lnnI56BMouqUOoOY13tscFo/atqTOGrbParqgmPzKM1YDD9r+QtP1RN/aDzgS+CHF8XZJua02bQ6LH0j6OrCWpLcBhwDfaLimmPxuLb8/1GQR0d/KUU/vlrS67Qd70WbrRkNJWtn2I+XjV1GsNyDgHNvnNVpc9B1Jz7L916briP4iaXvgm8DqtjeStAXwdtuH19ZmC8PiGttzJH3P9v9uup7oT5JeBnyLHv4wR3tIugLYBzjD9lbltuttb15Xm23shpoq6UBge0l7D96ZZS9jnHwR+HvgDADbv5C0U7MlRT+xfbuk7k2P19leG8PiHcD+FDex7DloX5a9jHHT6x/maJXby64oS5oK/CNPraVSi9aFhe1LgUslXW37W03XE32r5z/M0SrvAL4ErA8sAc4F3llng627ZtFN0ubAZsAqnW22j2+uougXkqZR/DC/kmIAxbnAu23XOhY+oi6tDYtyPPwuFGGxANgduNT2Pk3WFRGxPMtb272jzjXeW9cN1WUfYAvgWtsHS3oOxVC0iFFr8oc5WqGztvsOFH/ofr98/kZgoM6G2xwWD9l+QtIySWsCdwGbNF1UTHpXV78kYnRsfxdA0kHArrYfK58fS9HVWZs2h8XVktaiuGt7AHiQYrK3iFHr/DB3lH+I2PYDDZUU/el5wBpAZ/2K1ctttWntNYtukmYAa9pe1HAp0SckzQO+Q/EDLeA+4BDbtXYVRDuUSyscBVxYbtoZOGrwHyvj2mZbw0LSDsBC23+VdAAwB/iS7dsaLi36gKRFwDtt/3f5fEfgv2zPbray6BeSngt0Vvu8wvYf6myvtetZAMcAS8tpGD4I3AZk2GyMlwc6QQFP3t+TrqgYTysCd1Oslrdp3TMEtPmaxbJyivK9KM4ovlVOAxIxapLmlA+vLGc1PolidNS+wEVN1RX9RdJnKY6pG4Anys2dqcrrabPF3VAXAz8FDgZ2okjohbZf2mhhMalJunCY3ba9W8+Kib4l6SZgdmcG7V5o85nFvsBbgENt/0HSRsDnGq4pJjnbuzZdQ7TCzRRLQfcsLFp7ZhFRN0n/ALyEp08nc3RzFUW/kHQaxU3FF9AVGLmDexxJuoWib+9u29tWvT5iNMqbpFYDdqWYGWAfch9PjJ8zyq+eyZlFRA0kLbI9u+v76sDptl/ddG3RHyStCmxk+6ZetNfmobMRdeqswb1U0vOAx4CZDdYTfUTSnsBCikE6SNpSUq1nGgmLiHqcVU4n8zngGuBW4ORGK4p+chSwDcXMANheSM1/jLTumkVEj/x7OazxNElnUVzkfrjhmqJ/LLN9/6CVGGu9ppAzi4h6XNZ5YPsR2/d3b4sYo+slvQVYUdKscmr8n9fZYMKiJOl8ST+RtEfTtcTkJem5kuYCq0raStKc8msXitFREePhCIph2Y8AJwL3A++us8GMhiqVFyGnA9vZ/lrT9cTkVE4ZcxAwj6evbfEAcJzt05uoK/qLpDfaPqVq27i2mbCIGH+S3mD7tKbriP4k6Rrbc6q2jafWXeAu5+4xcG/W247xJukA2ycAMyS9d/B+219ooKzoE5J2B14LrC/py1271gSW1dl268KCoosA4PEmi4i+9azy++qNVhH96g6K7s3X8fQ1tx8A3lNnw63uhpK0MTDL9vnl3ZBTsvxlREx0klbqrL/dK208swBA0tuAw4B1gOcDGwDHAq9osq7oD5K+wxDj3m0f0kA50X+2kXQUsDHF73FRTIG/SV0NtjYsgHdS3AF5BYDt30j6u2ZLij5yVtfjVYDXU3QhRIyHb1F0Ow3Qoy71NofFI7Yf7dwBKWkKNd8BGe0xeCSUpJOA8xsqJ/rP/bZ/0ssG2xwWF0v6KMXNU68CDgfObLim6F+zgI2aLiL6xoWSPgecztPXs7imrgZbe4Fb0grAocCrKfr7zgG+6bb+h8S4kvQATz9T/QPwkdx7EeNhOcv31rpsb+vCQtIFtl8h6bO2P9R0PRERk0Ebu6GmS9oZeJ2kkynOKp5U52lctIekHYCFtv8q6QBgDvAl27c1XFpMYp2bPoe64RPqvemzjWHxceDDFENlB//HGqjtNC5a5RhgC0lbAB+kGL1yPLBzo1XFZNe56XONXjfcum6oDkn/YvuTTdcR/akzT4+kjwO/t/2tuufuiahTa8MCQNLaFKNUVulss31JcxVFv5B0McWSlwcDOwF3U3RLvbTRwmJSK//4AHiw1/OMtbEbCgBJ/4di/vcNKNay3Y5icZp0Q8V42Bd4C3Co7T9I2ohiidWIsehc83po2FfVoLVnFpKuA7YGLre9paQXAZ+wvW/DpUVETDitPbMAHrb9sCQkrWz7V5Je2HRRMblJuoVioMTdtrdtup6I8dL09+3DAAAIK0lEQVTmsFgiaS3gR8B5kv5M5u6JMbI9s+kaIurQ2m6obuV9F88Gfmr70abriYiYaFZouoCmSHq+pJU7T4EZwGrNVRQRMTqSDpe0bzkhai1aGxbAacDjkl5AccPUTODEZkuKiBgVATtSTCxYTwNt7YbqumnqAxQXu78i6VrbWzVdW0TERNPmC9yPSdoPOBDYs9y2UoP1RB+TdD7wGPA122dVvT5iKJLeWj58yPYpvWy7zWFxMPAO4NO2b5E0Ezih4Zqif70VmE5x82fEaHVG2z3Q64Zb2w0VEREj17oL3JIulPQzSac2XUv0nxxf0QuSNpV0gaTry+ezJX2s1jbbdmYhaePy4eO2lzRaTPSdHF/RC+VElR8Avt4ZlCPpetub19VmG69Z/K5q6VRJyvKqMRrdixuVwTHL9vmSVgWm2O55X3P0pdVsXyk9be22ZXU22LpuKIqFzo8oZwF9kqSpknaT9F2KEVIRoybpbcCpwNfLTRtQTC0TMR7ukfR8ynXeJe0D3Flng23shloFOATYn2JkwX0U61msCJxLMbRxYXMVRj+QtBDYBriiq5vguqxnEeNB0ibAfGB74M/ALcABtm+trc22hUU3SSsB0yjGLN/XdD3RPyRdYXvbzo2e5TQM19ie3XRt0T8kPQtYoRfdm228ZvEk249R86lbtNbFkj4KrCrpVcDhwJkN1xSTnKT3Lmc7AHWuntfqsIio0YeBQ4HrgLcDC4BvNlpR9IM1mmq41d1QEeNN0gW2XyHps7Y/1HQ9EeOljaOhIuo0vVwf5XWStpI0p/ur6eKiP0jaRNKZku6WdJekH5cXvetrM2cWEeOnHMJ4KMV00VcP2m3bu/W+qug3ki4HvgacVG56M3BEnUv5JiwiaiDpX2x/suk6oj91RtsN2na57domqkxYRNRE0trALIr7eACwfUlzFcVkJ2md8uEHKe4RO5nixrx9gZXr/AMlYRFRA0n/B3g3xZ3bCymmJr8s3VAxFpJuoQgHDbHbtmu7bpGwiKiBpOuArYHLbW8p6UXAJ2zv23BpEaOS+ywi6vGw7YclIWll27+S9MKmi4r+IWlzYDOe3s15fF3tJSwi6rFE0loUkweeJ+nPwB0N1xR9QtKRwC4UYbEA2B24FKgtLNINFVGz8r6LZwM/tf1o0/XE5Fd2c24BXGt7C0nPAb5pe8+62sxNeRE1kPR8SSt3ngIzgNWaqyj6zEO2nwCWSVoTuAuo9aa8hEVEPU4DHpf0AuBbFNPhn9hsSdFHri67Ob8BDADXAFfW2WC6oSJqIOka23MkfYDiYvdXOtOVN11b9BdJM4A1bS+qs52cWUTU4zFJ+1GsunhWuW2lBuuJPiLpgs5j27faXtS9rQ4ZDRVRj4OBdwCftn2LpJnACQ3XFJNcudLnasC0coaAzs15awLPq7XtdENFREwOkt4N/BNFMHQPxf4L8A3bX62t7YRFxPiRdCHFdAz32t6n6XqiP0k6wvZXetpmwiJi/EjauHz4uO0ljRYTfUfSbrZ/JmnvofbbPr2utnPNImJ8/c4Vf4FJUtVrIpZjZ+BnwFA33xmoLSxyZhExjiRdRHGPxY9t/65r+1SKBZEOBC60fVwjBUaMUsIiYhyVo1UOAfanuBHvPoqJ3lYEzgW+ZnthcxVGP5D08aG22z66tjYTFhH1kLQSMI1iaob7mq4n+oek93U9XQXYA7jR9iG1tZmwiIiY3Mp5yM6w/fd1tZE7uCMiJr/VqHkiwYyGioiYZMopyjvdQisC6wG1Xa+AdENFREw6XffzACwD/mh7WZ1tphsqImLymQL8wfZtwCzg8HLK8tokLCIiJp+er5eSsIiImHyeKLud9ga+aPs9wPQ6G0xYRERMPp31Ut5Kj9ZLSVhEREw+BwMvo4frpWQ0VEREVMp9FhERk0ST66XkzCIiYpJocr2UhEVExCQxkrVQ6lovJRe4IyImjwslHSFpo+6NkqZK2k3SdynWTBl3ObOIiJgkmlwvJWERETEJ9Xq9lIRFRERUyjWLiIiolLCIiIhKCYuIIUj6Z0k3SFokaaGkbWts6yJJ8+r6/IjxkDu4IwaR9DJgD2CO7UckTQOmNlxWRKNyZhHxt6YD99h+BMD2PbbvkPRxSVdJul7SfEmCJ88M/lPSJZJulLS1pNMl/UbSp8rXzJD0K0nfLc9WTpW02uCGJb1a0mWSrpF0iqTVy+3/JumX5Xs/38P/iwggYRExlHOBDSX9WtJ/Sdq53P5V21vb3hxYleLso+NR2zsBxwI/Bt4JbA4cJGnd8jUvBObbng38BTi8u9HyDOZjwCttzwGuBt4raR3g9cBLyvd+qoZ/c8SwEhYRg9h+EJgLHAbcDXxf0kHArpKukHQdsBvwkq63nVF+vw64wfad5ZnJzcCG5b7bbf9P+fgEYMdBTW8HbAb8j6SFFHfibkwRLA8D35S0N7B03P6xESOUaxYRQ7D9OHARcFEZDm8HZgPzbN8u6SiKO2c7Him/P9H1uPO883M2+Kamwc8FnGd7v8H1SNoGeAXwZuBdFGEV0TM5s4gYRNILJc3q2rQlcFP5+J7yOsJopofeqLx4DrAfcOmg/ZcDO5TrKiNpNUmblu092/YC4J/KeiJ6KmcWEX9rdeArktYClgGLKbqk7qPoZroVuGoUn3sjcKCkrwO/AY7p3mn77rK76yRJK5ebPwY8APy4nBdIwHtG0XbEmGS6j4gekDQDOKu8OB4x6aQbKiIiKuXMIiIiKuXMIiIiKiUsIiKiUsIiIiIqJSwiIqJSwiIiIiolLCIiotL/Byafynr+16JvAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x118151048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119d83860>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11a000710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11a01fd68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print (\"Text Classification some basics\")\n",
    "\n",
    "sep='+'*20\n",
    "import nltk\n",
    "from nltk.classify import apply_features\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import nltk\n",
    "import re\n",
    "\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.util import ngrams\n",
    "from nltk.probability import FreqDist\n",
    "from nltk.tokenize import word_tokenize\n",
    "\n",
    "path='data/structured_data.csv'\n",
    "strctured_df=pd.read_csv(path,error_bad_lines=False,sep='\\t')\n",
    "col_names=list(strctured_df)\n",
    "del strctured_df[col_names[0]]\n",
    "strctured_df.head()\n",
    "\n",
    "\n",
    "\n",
    "# Creating a smaller dataframe to play with a few features.\n",
    "# Data might be a bit disordered ?\n",
    "subset_df=strctured_df[['Overall rating for this service','Is the service caring?']]\n",
    "\n",
    "subset_df_grped=subset_df.groupby('Overall rating for this service')['Is the service caring?'].\\\n",
    "apply(lambda x: \"{%s}\" % ', '.join(x)).to_frame()\n",
    "subset_df_grped=subset_df_grped.reset_index()\n",
    "subset_df_grped\n",
    "\n",
    "\n",
    "num_rows=subset_df_grped.shape[0]\n",
    "first_n_common=10\n",
    "fetr_list=[]\n",
    "for index in range(num_rows):\n",
    "    words=[]\n",
    "    filtered_words=[]\n",
    "    #text=subset_df['Is the service caring?'][index].lower()\n",
    "    #tagged_sentence = nltk.tag.pos_tag(text.split())\n",
    "    text=subset_df_grped['Is the service caring?'][index].lower()\n",
    "    tagged_sentence = nltk.tag.pos_tag(text.split())\n",
    "    edited_sentence = [word for word,tag in tagged_sentence \n",
    "                       # To only collect adverbs and adjectives... \n",
    "                       if tag in ('JJ','RB')]\n",
    "                       # To remove the Nouns... \n",
    "                       #if tag not in ('NN', 'NNP', 'NNPS', 'NNS')]\n",
    "    #edited_sentence\n",
    "    # Just ensure the sentence has alphabetical characters \n",
    "    words = [word for word in edited_sentence if word.isalpha()]\n",
    "    \n",
    "    # We can also remove the stopwords..the, in, to, that...\n",
    "    #filtered_words = [word for word in words if word not in stopwords.words('english')]\n",
    "    filtered_words = words\n",
    "    \n",
    "    # Change the number of words in a pairs\n",
    "    fdist = FreqDist(ngrams(filtered_words,3))\n",
    "    \n",
    "    most_imp = FreqDist(dict(fdist.most_common()[:first_n_common]))\n",
    "    fetr_list.append(most_imp)\n",
    "keys=subset_df_grped['Overall rating for this service'].tolist()\n",
    "dictionary = dict(zip(keys, fetr_list))\n",
    "lst_dic=list(dictionary.keys())\n",
    "for index in range(len(lst_dic)):\n",
    "    dictionary.get(lst_dic[index]).plot(title=lst_dic[index])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}