A retrieval based chatbot based on NLTK library and trained using MaxEnt classifier
- Go to any FAQ website and copy all of the questions & answers there into a text file
- Or use one of the files in the repo
- Change the FILENAME constant to this file and run
- Subject a dataset (NPS Chats in our case) to NLP transformations like stemmatization, lemmatization, tokenization etc.
- Use this to train a classifier (Try both Naive Bayes and MaxEnt)
- Scrape Q&A text from the web and paste it into .txt file
- Use the classifier to separate questions from the text
- Add questions into a dictionary as keys and their answers as values
- Accept user input and apply the same transformations as in step 1
- Match user query with the question in the dictionary using TF-IDF
- Display the corresponding answer