In the work from home scenario prevailing over the last 18 months and with hybrid working picking steam, most Official meetings have been or will be conducted virtually. While MICROSOFT TEAMS / Google Meet / Zoom do provide a feature to download TRANSCRIPT, it does not summarise the meeting.
Problem Dimensions
- Parse the Transcript to figure out many attendees were there in the meeting.
- Duration of the meeting
- Most importantly produce a gist / summary of the meeting
- List the Action items if they are specifically called out.
- Incase where Transcript is not available, you will have to additionally work on converting SPEECH to TEXT
Project features Implementated are
- Speech to text
- Text summarization
- Action items
- Meeting duration
- Meeting attendees Many More ...
- Python 3.6+
- Good GPU (Nvidia GTX 1080 or better) to run all models in parallel
git clone https://github.com/gamingflexer/meeting-summarization-api.git
cd src/main-api
pip3 install -r requirements.txt
python manage.py makemigrations
python manage.py migrate
API SERVER
python manage.py runserver
Chatbot
python manage.py tailwind init
python manage.py tailwind install
python manage.py runserver 0.0.0.0:8001
redis-server
python manage.py tailwind start
cd src/models-api
python __init__.py
Some extra steps are required to run the application in production and development. run this inside models-api
folder
pip install git+https://github.com/openai/whisper.git --no-deps
python -m spacy download en_core_web_sm
pip install -U scikit-learn scipy matplotlib
git clone https://github.com/maxent-ai/converse.git && cd converse && pip install -e . && cd ..
sudo apt install ffmpeg
sudo apt-get install libsndfile1-dev
PermissionError:
sudo chmod -R 777 path of the media folder`