Treffer: Navigating the Evolving Landscape of Multimodal Sentiment Analysis: Recent Advances and Insights

This survey study offers a thorough introduction to multimodal sentiment analysis (MSA), emphasizing the advancements and challenges faced when combining various data modalities, including text, audio, and visual inputs, to improve sentiment prediction. Data from a variety of sources, such as social media, movies, and customer service encounters, as well as crowdsourcing and expert labeling, are all examined in this research. Numerous approaches covered in the literature are examined, with an emphasis on sophisticated models such Recurrent Multimodal Sentiment Analysis, Multimodal Graph Convolutional Networks (GCNs), VisualBERT, ViLBERT, Multimodal Fusion Transformer (MFT), and Deep Multimodal Sentiment Analysis (DMSA). These models better perceive and predict sentiment across various media by combining many neural network designs and attention mechanisms. The study also covers performance indicators that are frequently used to assess these models. It identifies key challenges and suggests future enhancements to improve the scalability, efficiency, and accuracy of multimodal sentiment analysis systems.