Researchers have developed a new zero-shot stance detection framework called KIRP, designed to improve the accuracy of identifying stances in short texts like tweets. The framework addresses challenges such as context sparsity and relevance between implicit targets and content by integrating external knowledge graphs and employing reflective Chain-of-Thought (CoT) reasoning. KIRP also utilizes stance-aware contrastive learning and a three-layer iterative prototype network to better distinguish between neutral and irrelevant labels. Experiments on several datasets, including a newly constructed Japanese tweet dataset, show that KIRP achieves state-of-the-art performance with high F1 scores. AI
IMPACT This research could improve the accuracy of AI systems in understanding nuanced opinions and sentiments expressed in short-form text, impacting social media analysis and content moderation.
RANK_REASON The cluster describes a new research paper detailing a novel framework and dataset for stance detection.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
