Text-Based Emotion Recognition: A Review

As the world is becoming more modern and digital, the interaction of humans with computers is a very intriguing and well-known subject of research. Computer interfaces must correctly detect users’ emotions to develop highly intelligent behavior. In recent years, emotion recognition through text has been studied in various disciplines like machine learning (ML) and natural language processing (NLP). The main objective of the study is to provide a comprehensive analysis through a thorough discussion and overview of various research trends in text-based emotion recognition and prediction systems. Various methodologies were tested on various datasets in the reviewed papers under the domains of ML and NLP. We analyzed that the majority of research is conducted using machine learning techniques like hybrid method (keyword-based + learning-based), lexical affinity, Nave Bayes, support vector machine (SVM), long short-term memory (LSTM), etc. The results of this study highlight the continuous growth in this research area. Moreover, from a comprehensive study of highly cited publications, we retrieved those algorithms like multinomial Naive Bayes, LSTM, and K-means significantly provided high accuracy irrespective of the complexity of the text.