Interpretable Sentiment Analysis based on Deep Learning: An overview

Abstract

Sentiment analysis (SA) or emotion AI or opinion mining uses natural language processing (NLP). Sentiment Analysis identify, study, quantify, obtain, tacit states and subject related information. Broad spectrum of areas influenced due to Sentiment Analysis such as policy making by the government, finding mental health of individuals, finding misuse of drugs in healthcare, fraud detection in the financial sector, covid-19 awareness and impact, Cyber-crime etc. As the amplitude of social media data increases day by day, there is a need to automatically address sentiment analysis. Deep learning handles it very well. It gives very good accuracy but incomprehensibility in decision strategy. For better decision-making trust, believe, fairness, reliability, and unbiasing is important. This paper explores the work done in this area along with popular techniques to address interpretability in sentiment analysis and its evaluation criteria.