A Review on Sentiment Analysis from Multimodal Data

Abstract

A sentiment analysis is becoming a popular research area.  Sentiments can be expressed in various forms like text, image, audio, video etc. Opinion of large population is anticipated by aggregating sentiments of individual and used in numerous applications. Traditional sentiment analysis system focuses on single modality to infer user’s perception about subject. Such type of sentiment analysis has its own limitation and fails to employ other modality’s expressiveness.

 With the advent of technologies, the conventional system evolved into multimodal sentiment analysis which integrates multifarious data (i.e. text, audio, and visual) available over internet. Multimodality refers to the availability of more than one modality or medium.  Every mode of data has unique features and helps users to express their emotions, opinions or attitude about the entity. Incorporating such features from multiple content enhance the effectiveness of sentiment analysis process. To find a constructive fusion mechanism to integrate these features is the challenging aspect of sentiment analysis.  In this survey we have defined different modalities of sentiments, characteristics and fusion techniques of multimodal data.  This paper gives an overview of different approaches for and applications of multimodal system.