Instruction tuned LLMs identify argument parts directly from text
Extracting Arguments, Not Just Classifying Them: Instruction-Tuned LLMs for Generative Component Detection
Artificial IntelligenceComputation and Language
Summary
Figuring out the parts of an argument in a text, like claims and reasons, is tricky. The authors presented a new method using large language models that follow clear instructions to find and classify these parts without first chopping the text into segments. Their approach treats this task as generating text rather than just labeling segments. Tests show this method works better than previous ones on well-known datasets. This shows new ways to use instruction-guided language models for understanding arguments.
What this means in practice
- •For legal document analysis teams: Automatically identify and classify argument parts in legal texts to aid document review and case preparation.
- •For social media moderation teams: Detect argumentative components in user posts to better understand the structure of debates and flag misleading claims.
Authors
Sofiane Elguendouze, Erwan Hain, Elena Cabrio, Serena Villata
Abstract
Argumentative component detection (ACD) is a core subtask of Argument(ation) Mining (AM) and one of its most challenging aspects, as it requires jointly delimiting argumentative spans and classifying them into components such as claims and premises. While research on this subtask remains relatively limited compared to other AM tasks, most existing approaches formulate it as a simplified sequence labeling problem, component classification, or a pipeline of component segmentation followed by classification. In this paper, we propose ITFACD, a novel approach based on instruction-tuned Large Language Models (LLMs) using compact instruction-based prompts, and reframe ACD as a language generation task, enabling arguments to be identified directly from plain text without relying on pre-segmented components. Experiments on standard benchmarks show that our approach achieves higher performance compared to state-of-the-art systems. To the best of our knowledge, this is one of the first attempts to fully model ACD as a generative task, highlighting the potential of instruction tuning for complex AM problems. Our code and the datasets used are openly available in the following GitHub repository.