Getting answer automatically is magic!! its real AI (remember, the Turing Test?)

This project is a Simple Question-Answer (atomic query) based chatbot framework. Uses similarity based on different vectorizers, to find the matching question then responds with its corresponding answer.

Application Scope:

• Huge demand to take care of mundane queries
• Scales (leverage, automation, passive)
• Not much work in vernacular chatbot (serve humanity)

Notes:

• This chatbot is based on category classification first and then to similarity within the selected category.
• Different than the popular open source chatbot framework, Rasa, where NLU is based on intent and entities, whereas dialog management is based on sequence/LSTM prediction.
• Conceptually it is similar to Microsoft's QnA Maker. But the big difference is that, if you get whole this whole github code-base, your models would be local. Nothing on Server. So better security especially for sensitive data chatbots like HR or Finance.

Copyright (C) 2019 Yogesh H Kulkarni

• Implement sentence embeddings via HuggingFace or Spacy
• Build full FAQ chatbot platform using switchable embediddings
• Deploy it on Heroku and give link from Yati.io with documentation and disclaimer
• Research: SIG IR: entity extraction, Answer selection is part of Information retrieval

• You supply FAQs in the form of csv (comma separated file) having Question-Answer-Class in each row (e.g. "What is GST rate for Toothpaste?,12,rate")
• Questions are vectorized and kept ready for matching, along with the classifier model [X=vector(question), y=class]
• Once user query comes, its 'class' is predicted using the classifier model and within the class, vectorized query is matched against existing vectorized questions.
• Whichever is most similar, it's answer is presented to the user.

• app.py: Chatbot UI built using Flask, using templates/*.html
• bankfaqs.py: Chatbot core logic as well as knowledge-base.

• faqs: csv files containing questions and answers
• static and templates: Flask UI related files

• Execute app.py
• Open http://127.0.0.1:8080/ in the browser
• Start chating

• Needs Python 3.6, numpy, scipy, sklearn

• Bhavani Ravi’s event-bot code, Youtube Video
• Banking FAQ Bot code

• Author ([email protected]) gives no guarantee of the results of the program. It is just a fun script. Lot of improvements are still to be made. So, don’t depend on it at all.