arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Artificial Intelligence

Multimodal Sentiment Analysis Using Pre-trained Models: A Study of Few-Shot learning and Social Media Scenarios

F.C. Lin, Junhao Yu, Ziqi Ye

· Applied and Computational Engineering

See on the map

Abstract

Facing the proliferation of social media and user-published content, traditional single sentiment analysis cannot capture users' emotional expressions well, and multi-modal sentiment analysis (MSA) can solve this problem well. Multi-modal sentiment analysis improves the system's ability to understand complex emotional data through the integration of multi-modal tasks such as audio, image, and video. This paper systematically analyzes five representative multi-modal sentiment analysis pre-training models in recent years, focusing on the model structure, integration methods, adaptability and performance of these five models in small sample-oriented and social media scenarios. Through the analysis and summary of MGT, SensoryT5, AMP, MMEN and VLP2MSA models, these models have shown good robustness and generalization in complex emotional tasks with small samples, especially in modal alignment, feature fusion, and fine-grained emotion recognition. Significantly improved. This paper provides a systematic reference for the selection and optimization of pre-trained models for multi-modal sentiment analysis, all of which promote the improvement of the robustness and generalization of the model in the scenarios of small samples, social media, and complex emotions.

More like this