Sentiment analysis and emotion recognition in social media: A comprehensive survey

Abstract

Sentiment Analysis (SA) and emotion recognition is the fundamental dialogue system that recently gained more attention. It is applied in many scenarios like mining the opinions of the speaker’s conversation and enhancing the feedback of the robot agent. Furthermore, the live conversation is used to generate the talks through certain sentiments to enhance the human-machine interaction. This survey focuses the researchers on handling the SA and classification of various sentences in social media by reviewing various approaches. This analysis explains the 50 research articles from different methods used for SA and sentiment classification in social media. Finally, the evaluation of this survey is performed based on the publication year, various approaches, evaluation metrics, and tools. Moreover, the collected 50 research papers are categorized into different techniques, such as deep learning (DL) based methods, machine learning (ML) based methods, lexicon-based methods, hybrid-based methods, and dependency-based methods. Therefore, from this survey, it is clearly shown that the DL-based method is the most utilized approach in many research papers. Similarly, python is the most used tool for SA and classification, and real-time dataset is a commonly used dataset for SA and classification. Likewise, accuracy is repeatedly employed in metrics with the highest value.