Emotionally Intelligent ChatBot for Mental Healthcare and Suicide Prevention

Abstract

Mental Health is an integral part of the human body. People suffer from various mental health problems. Direct Counseling is not affordable for many, while some are not open to express their inner thoughts to another person. We aim to use “Cognitive behavioral therapy” which uses psycho-social intervention that aims to improve mental health. To provide a free service of interaction with a machine, the objective of this “Emotionally Aware Chatbots” named as “SA-DO” is to provide mental healthcare to those who are mentally-ill anywhere and anytime. Methodologies used in this project are of Natural Language Processing focusing on “Dialogue Generation” by understanding the context of the text and “Sentiment Analysis” to detect the mental state of the person and respond accordingly by providing apt advice and analysis. If we can detect mental health issues at the correct time and provide suitable help, we might save precious lives. To achieve a high accuracy in sentiment classification, a “Logistic Regression” is used which uses the concept of “tfidf” to do multi-classification of sentiments in the dataset which gives an accuracy of 89%. NLP techniques using regular expressions are used for intent classification and entity extraction. In this project, we have also tried comparing the performance of the two algorithms which are Logistic Regression and the other one Random Forest Classifier.

  Keywords:   Machine Learning, Artificial Intelligence, Natural Language Processing, Entity  Extraction, Intent Classification, Sentiment Analysis.