This study investigates whether merging abstract logical structures with context-level linguistic cues improves the automated classification of logical fallacies, which often appear in nuanced forms.
The authors developed a framework that inductively extracts these patterns from fallacious examples and their explanations using Large Language Models (LLMs).
Experiments across different LLMs and zero- and one-shot configurations demonstrated statistically significant improvements over zero-shot baselines and outperformed competing approaches.
Cross-dataset experiments validated the generalization of this method, establishing data-driven pattern extraction as an effective approach for generating logical representations.