Abstract, Align, Predict: Zero-Shot Stance Detection via Cognitive Inductive Reasoning

2025-06-16

AI Summary

This paper addresses zero-shot stance detection—identifying textual stance toward unseen targets without target-specific labeled data—by proposing the Cognitive Inductive Reasoning Framework (CIRF). CIRF models human abstract reasoning at the conceptual-logical level and introduces the Schema-Enhanced Graph Kernel Model (SEGKM), which dynamically aligns local reasoning patterns with global structural knowledge. Fully unsupervised, CIRF requires no target-domain annotations, achieving cross-target generalization solely through reasoning-pattern abstraction and conceptual-logical encoding. Evaluated on SemEval-2016, VAST, and COVID-19-Stance benchmarks, CIRF achieves new state-of-the-art macro-F1 scores, improving by 1.0, 4.5, and 3.3 percentage points, respectively. Remarkably, using only 30% of labeled training data, CIRF matches the performance of fully supervised models.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Knowledge Representation and Reasoning: Argumentation
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Zero-shot stance detection (ZSSD) aims to identify the stance of text toward previously unseen targets, a setting where conventional supervised models often fail due to reliance on labeled data and shallow lexical cues. Inspired by human cognitive reasoning, we propose the Cognitive Inductive Reasoning Framework (CIRF), which abstracts transferable reasoning schemas from unlabeled text and encodes them as concept-level logic. To integrate these schemas with input arguments, we introduce a Schema-Enhanced Graph Kernel Model (SEGKM) that dynamically aligns local and global reasoning structures. Experiments on SemEval-2016, VAST, and COVID-19-Stance benchmarks show that CIRF establishes new state-of-the-art results, outperforming strong ZSSD baselines by 1.0, 4.5, and 3.3 percentage points in macro-F1, respectively, and achieving comparable accuracy with 70% fewer labeled examples. We will release the full code upon publication.

Problem

• Identify stance toward unseen targets without labeled data
• Transfer reasoning schemas from unlabeled text to logic
• Align local and global reasoning structures dynamically

Innovation

• Cognitive Inductive Reasoning Framework for ZSSD
• Schema-Enhanced Graph Kernel Model integration
• Dynamic alignment of reasoning structures