The field of emotion understanding and human behavior analysis is moving towards more nuanced and multimodal approaches, incorporating large language models and fusion of different modalities to better capture subtle emotional cues and human behavior. Recent developments have focused on addressing challenges such as the entanglement of static and dynamic cues, semantic gaps between text and physical motion, and the need for more fine-grained multimodal fusion strategies. Noteworthy papers include DEFT-LLM, which achieves motion semantic alignment by multi-expert disentanglement, and MemoDetector, which introduces a dual-stage modal fusion strategy to better capture nuanced cross-modal emotional cues. Other notable works include GazeInterpreter, which parses eye gaze data to generate eye-body-coordinated narrations, and Unveiling Intrinsic Dimension of Texts, which establishes a comprehensive study grounding intrinsic dimension in interpretable text properties.
Emotion Understanding and Human Behavior Analysis
Sources
DEFT-LLM: Disentangled Expert Feature Tuning for Micro-Expression Recognition
Enhancing Meme Emotion Understanding with Multi-Level Modality Enhancement and Dual-Stage Modal Fusion
Synergistic Feature Fusion for Latent Lyrical Classification: A Gated Deep Learning Architecture
Classification of Hope in Textual Data using Transformer-Based Models
Rdgai: Classifying transcriptional changes using Large Language Models with a test case from an Arabic Gospel tradition
Encoding and Understanding Astrophysical Information in Large Language Model-Generated Summaries
Unveiling Intrinsic Dimension of Texts: from Academic Abstract to Creative Story
GazeInterpreter: Parsing Eye Gaze to Generate Eye-Body-Coordinated Narrations
Beyond Tokens in Language Models: Interpreting Activations through Text Genre Chunks