Novel approaches for multimodal sentiment analysis using deep learning

Abstract

Every day, a lot of videos are added to social media sites like Facebook and YouTube. As a result, the web becomes a source of information that never runs out. In the next few decades, it will become more and more difficult to deal with this kind of information and get useful information out of it. To this point, the majority of research in sentiment analysis has been done on natural language processing. People can use text-based data and resources to study how people feel about things. People have been using social media to share their thoughts. Social media platforms are becoming more and more popular places for people to share videos, photos, and audios with each other. People are increasingly using these types of media to express themselves on these platforms. As a result, it is important to look for opinions and feelings from different ways. The main objective of multimodal sentiment analysis is to gather and aggregates the sentimental information related to the customer spawned multimodality information. newlineIn this thesis, the general problem of predicting sentiment is studied based on various modality. A comprehensive survey is presented which investigates the sentiment analysis concepts, identifies the roles of extracting relevant information from different modality such as text, audio newline