A Comprehensive Review of Multimodal Sentiment Analysis on Social Networks

In the realm of emerging technologies, there is a scientific discipline dedicated to enabling expert systems to not only detect and predict but also comprehend human emotional responses. The application of the fast-evolving discipline of natural language processing known as multimodal sentiment analysis, is gaining more importance. It makes an effort to determine the opinions voiced by users by gathering and synthesizing semantic data obtained from various modalities. The results of predictions are affected in real time by the complimentary and different information among the various modalities. Recent work creates independent representations of all modalities using a single neural network. Every modality offers a unique set of benefits that can be combined to ascertain whether consumers have positive or negative opinions. The most recent advancement in multimodal sentiment analysis combines auditory, visual, and textual modalities to increase accuracy. In this paper describes using the most recent datasets, a variety of sentiment analysis fusion methods that leverage on various modalities have been produced, and the challenges faced are discussed.