Research on Documentary Narrative Optimization Based on Multimodal Sentiment Analysis
Xiaoya Dai, Yanni Wu, Jiake Li · 2025
This research explores the application of deep learning-based emotion analysis technology in documentary production. The paper proposes a multimodal emotion analysis framework that combines Transformer models to identify and quantify emotions in documentary text, visual, and audio content, thereby helping creators with material selection, emotional rhythm control, and narrative optimization. Through experimental validation on the Go Emotions dataset, the proposed model demonstrates superior performance in emotion recognition tasks, particularly in handling complex emotional expressions and non-verbal cues.