Papers by Tiantian Liang
Why Do Emotions Change? Appraisal-Guided Reasoning for Emotion–Cause Triplet Extraction in Conversations (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for multi-turn, multi-speaker multimodal affect understanding are difficult to maintain conversation-level consistency under within-speaks' emotion shifts. |
| Approach: | They propose a framework that combines appraisal-guided structured generation with graph-structured reinforcement learning to extract triplets from multi-turn multimodal conversations. |
| Outcome: | The proposed framework outperforms baselines on public MECTEC benchmarks and improves structure-aware metrics on emotion shift coherence and core events. |
M3HG: Multimodal, Multi-scale, and Multi-type Node Heterogeneous Graph for Emotion Cause Triplet Extraction in Conversations (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for ECAC focus on textual contexts, overlooking other modalities. |
| Approach: | They propose a multimodal, multi-scenario MECTEC dataset that captures emotional and causal contexts and effectively fuses contextual information at different levels. |
| Outcome: | The proposed model captures emotional and causal contexts and effectively fuses contextual information at both inter- and intra-utterance levels. |
Multi-Graph Co-Training for Capturing User Intent in Session-based Recommendation (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods rely on user actions within the current session, overlooking the wealth of auxiliary information available. |
| Approach: | They propose a session-based recommendation model that leverages the current session graph and similar session graphs to capture the intrinsic relationships between items. |
| Outcome: | The proposed model improves on the Diginetica dataset by 2.00% and 10.70% respectively. |