Zero-shot Tweet-Level Stance Detection Enhanced by External Knowledge and Reflective Chain-of-Thought Reasoning

Researchers propose KIRP, a zero-shot stance detection framework that addresses context sparsity and implicit target relevance in short texts by integrating external knowledge with reflective Chain-of-Thought reasoning. The study also introduces the first Japanese tweet-level dataset for stance detection to support this multi-topic evaluation.

• KIRP utilizes knowledge graphs to supplement and reorganize key textual entities for data augmentation.
• Reflective Chain-of-Thought reasoning is employed to extract and validate implicit targets within the text.
• Stance-aware contrastive learning and a three-layer iterative prototype network distinguish between "neutral" and "irrelevant" labels.
• The framework achieves state-of-the-art performance with F1 scores of 84.05% on SemEval-2016, 84.99% on WT-WT, and 79.18% on the new KIRP-D dataset.