Curious about the texting habits of you and your friends? Look no further! Chatty will show you stats on how you communicate with whatever individual you tell it to analyze. Out of the box, Chatty supports iMessage, WhatsApp, and Line. Custom sources and arbitrary text is supported as well.
Why not? Texting styles are a hot topic of discussion in my friend group. I wanted quantitative evidence that some of my friends are better texters than others. Kidding. Kind of.
Please see the project notebook for a full demonstration. Note that GitHub does not render Jupyter notebooks on mobile. If you "Request as Desktop Site" using your mobile broswer, it's still fairly readable.
Short of that, two examples (out of the many in the notebook) are a comparison table generated by summary_comparison:
| Metric | From Me | To Me |
|---|---|---|
| Message Count | 1,199 | 960 |
| Word Count | 29,686 | 14,243 |
| Avg Message Word Count | 24.76 | 14.84 |
| Avg Message Word Count (w/o Stop Words) | 14.48 | 9.36 |
| Avg Message Character Count | 96.38 | 55.48 |
| Avg Message Character Count (w/o Stop Words) | 68.51 | 41.03 |
| Longest Message | 208 | 91 |
| Distinct Words | 4,345 | 2,631 |
| Distinct Stems | 3,438 | 2,229 |
| Distinct English Words | 3,773 | 2,248 |
| Distinct Emojis | 38 | 32 |
| Emoji Count | 193 | 118 |
| Pct Negative Sentiment | 7% | 4% |
| Pct Positive Sentiment | 18% | 19% |
and sentiment analysis carried out and displayed by sentiment and sentiment_plot:
A working Python 3 Anaconda environment is required. Once Anaconda is installed, run:
conda env create -f environment.yml
to install Chatty's dependencies.
Launch chatty by running:
source activate chatty
jupyter notebook
Select a source per the instructions below, and re-evaluate all cells.
Chatty includes a utility for reading config from an external yaml file: chatty.load_vars(vars_file='vars.yml'). Use this method for situations where you wish to share your notebook without exposing sensitive variables (e.g. names or telephone numbers).
The iMessage source requires an OSX computer with iMessage. Once the notebook loads, redefine RECIPIENT_ID to either an E.164 number of the form '+13031234567' or an iMessage handle of the form '[email protected]'.
The WhatsApp source reads from a WhatsApp chat log exported by the iOS WhatsApp client. It has not been tested against other clients.
from chatty.sources import whatsapp_chat
project_vars = load_vars()
all_messages = whatsapp_chat(project_vars['chat_path'])
from_messages = all_messages[project_vars['from']]
to_messages = all_messages[project_vars['to']]The line source reads from a Line chat log exported by the iOS Line client. It has not been tested against other clients.
from chatty.sources import line_chat
project_vars = load_vars()
all_messages = line_chat(project_vars['chat_path'])
from_messages = all_messages[project_vars['from']]
to_messages = all_messages[project_vars['to']]All analysis functions are composable and built for reuse. By default, Chatty will analyze messages sent to/from the counter party defined by RECIPIENT_ID. You can change this behavior by redefining from_messages and to_messages before the analysis functions are called.
- Support other message sources
- Add time series/trend analysis
- PRs welcome
Chatty is released under the MIT License.