Designing an Emotionally Realistic Chatbot Using Deep Learning

Abstract

Emotions are extremely important in human contact. With the exponential growth of information and communication technology over the last decade, chatbots are becoming an increasingly popular choice for interacting with users in most industries. Deep learning has recently gained popularity in various industries, and it can be used to solve the issues in developing emotionally realistic chatbots. The study introduces an emotionally realistic chatbot using deep learning based on Natural Language Processing (NLP) and Long Short-Term Memory (LSTM). To identify eight emotions from user inputs, the study employed a dataset called “Emotion Dataset for Emotion Recognition Tasks” along with a self-made dataset. The chatbot has been developed using TensorFlow and a deep learning model. The chatbot model consisted of two hidden layers and a typical feed-forward neural network. By using LSTM-based text emotion categorization, the chatbot delivers a higher level of accuracy than the existing learning methods. The results reveal that the system with emotion characteristics produces a statistically significant improvement in chatbot credibility when compared to the system without emotion variables. This effort could be utilized to create chatbots that can recognize a wide spectrum of emotions and be used for more critical activities, such as mental health care assistants. In the future, the model will be rewarded for emoji responses that are relevant to the sentiment. This will entail the use of Bi-LSTM, which provides the highest level of accuracy.