A Novel Semantic Approach for Intelligent Response Generation using Emotion Detection Incorporating NPMI Measure

Abstract

Expressions and emotions are the most common way of communication in day-to-day life. In the age of Artificial Intelligence and technological advancements, the entire human race finds itself amidst many software driven voice-assistants. The only reason AI cannot excel and spread its limits is that humans can interpret, understand and express in the form of emotions and these AI-driven systems cannot. Hence, there is a need for a proper methodology for the interpretation of emotions based on both text and speech. In order to accomplish this task, a light weight computational linguistic semantic approach has been proposed for detecting emotions and generating response incorporating NPMI and NAVA words, bridging the gap between Semantics and Natural Language Processing. Experimentations are conducted for the real-word TDIL dataset for emotions such as joy, sorrow, anger, disgust, and fear. The proposed approach yields an accuracy of 96.155% for the emotion joy and 82.44 % for fear which definitely is the best-in-class accuracy for such systems.