Search

Group sentiment analysis in today's era is influencing many applications. Sentiment recognition is applicable for individuals, and group as well as crowd. This paper aims to highlight the ongoing present research scenario on perceived group sentiment analysis. The paper provides a brief overview of the strengths and weaknesses of well-known group sentiment analysis approaches. These approaches are broadly split into two basic categories: top-down and bottom-up approaches. A combination of top-down and bottom-up approaches is found superior in performance. This comprehensive review also discusses the available dataset with taxonomic representation and the importance of fusion techniques in group emotion recognition. This paper also narrates the novel attempt to demarcate the group emotion recognition approaches into the categories evolved by the influence of machine learning and deep learning at the feature extraction level and at the classification level. The multi-modality based group sentiment analysis has a great scope to overcome the challenges and improve the performance on real time applications. • Comprehensive survey on perceived group sentiment analysis. • Taxonomic representation of available group emotion datasets. • Use of various attributes which utmost affect a person's emotion in a crowd or group. • Use of fusion methods at feature and decision level in GER. • Performance comparison of the existing methodology. [ABSTRACT FROM AUTHOR]