KHEC CHATBOT

Abstract

A Chabot is something that clears your all doubt and queries in real time based without operator or any administrator continuous involvement all the way using natural language. Technically, it is a way of training a model that learns predefined queries and its corresponding response so that queries, Chabot get through any input device mainly either keyboard or microphone is tested against trained dataset which result of its corresponding response and provide relevant response in short interval of time. Clearly, the aim is to clear the response in real time and quick way. So as with same aim, not only clearing queries we proposed an interactive and clean sheet Chabot based on text as well as speech to make conversation more alive. The text model goes through pipeline that involves contractions, tokenization, cleaning, Bag of Words and last but not the least probability calculation. To keep clean sheet we used spam and block concept which adds on security concept. Believing on keeping conversation more alive we integrated speech model and develop two model where one implement classification model and next one involved reusing the text model pipeline and totally input/output on speech based where first one involved only input in speech based which proceed by extracting features and its corresponding class using MFCC concept and sequential classification model. This design would make more interactive and natural with less effort and helps to keep sanitized.