A Comprehensive Survey on Multimodal Sentiment Analysis

Sentiment analysis is a hot topic in the field of computer vision, natural language processing, and artificial intelligence. Nowadays, people share their reviews on products and services not only through text but also via audio, video, and images. The growing need for sentiment analysis across a wide range of channels of expression has made multimodal sentiment analysis (MSA) a prominent area of study. This survey paper categorizes various MSA fusion architectures into ten distinct categories. Within each category, recent contributions are identified and discussed, highlighting their strengths and limitations. Finally, numerous applications, challenges, and future directions are explored.