Khalil Mrini, Marc Laperrouza, Pierre Dillenbourg
Best Presentation Award at SwissText 2018 (Winterthur, Switzerland): Paper, Presentation.
We build a conversational agent which knowledge base is an online forum for parents of autistic children. We collect about 35,000 threads totalling some 600,000 replies, and label 1% of them for usefulness using Amazon Mechanical Turk. We train a Random Forest Classifier using sent2vec features to label the remaining thread replies. Then, we use word2vec to match user queries conceptually with a thread, and then a reply with a predefined context window.
The jupyter notebook file details the process to build the chatbot. The README
file in the Chatbot folder details how to launch the Django-based interface to chat with the Chatbot. However, pickle files are needed for it to run, that are too large to put on GitHub. Please contact the author to get them. You will also need the word2vec model GoogleNews-vectors-negative300.bin
.
After getting the files, run in the Chatbot
folder:
pip install -r requirements.txt
python manage.py runserver