This project is about the creation of a French-speaking chatbot (virtual assistant) that answers user questions about Coronavirus. This chatbot implemented in Python uses a Deep Learning library, PyTorch, giving us the tools to build a neural network model that is required to make complex classification.
The neural network is based on training data provided in a JSON file. This file is composed of different intents: Each intent gathers a few patterns (questions) written by users about differents topics focused on the Coronavirus (what is the coronavirus, its symptoms, how to wear a mask, etc), and also the associated label (or tag) for each pattern. This allows us to train the neural network model in a supervised manner. A list of responses is also available to make our chatbot able to speak and reply to users.
As of now, the intents our French Chatbot can understand are the following : aide psychologique, au revoir, coronavirus, couvre-feu, dépistage, déplacement, gestes barrières, merci, port du masque, salutation, symptômes.
A simple version of the app can be run on CLI via the following procedure:
pip install -r requirements.txt
python cli-chat.py
Afterward, a more sophisticated version of the app has been built using Flask framework and is now available on Heroku : https://fr-coronavirus-chatbot.herokuapp.com/