Unified Argument Mining Pipeline as Text Generation with LLMs

ACL ARR 2025 July Submission603 Authors

Abstract: In Natural Language Processing (NLP), Argument Mining (AM) focuses on identifying and extracting the underlying argumentative structure of a text. An AM pipeline takes a text as input and outputs its argumentative structure by successively classifying the arguments within the text (ACC task), identifying the argumentative relations between the arguments (ARI task), and classifying these identified relations (ARC task). Modern LLM-based approaches to AM reformulate the pipeline sub-tasks individually, converting them from text classification to text generation through innovative prompting techniques. In this work, we propose a novel LLM-based approach that addresses the successive sub-tasks of the AM pipeline as a unified text generation task, instead of treating them independently. We introduce an efficient 2-step prompting strategy that instructs the LLM to solve the ACC task and the joint ARI+ARC task in a single inference pass. Our unified AM pipeline approach achieves competitive or state-of-the-art performance on AAEC, AbsRCT, and CDCP benchmark datasets, outperforming or matching the best existing methods in which LLMs are fine-tuned separately for each sub-task. Overall, our work establishes `AM pipeline as text generation' as a rigorous and efficient AM paradigm and builds strong baseline results for future research.

Paper Type: Long

Research Area: Sentiment Analysis, Stylistic Analysis, and Argument Mining

Research Area Keywords: Argument Mining, LLMs, Fine-Tuning, AM Pipeline, Prompting

Contribution Types: NLP engineering experiment, Publicly available software and/or pre-trained models

Languages Studied: English

Reassignment Request Area Chair: This is not a resubmission

Reassignment Request Reviewers: This is not a resubmission

A1 Limitations Section: This paper has a limitations section.

A2 Potential Risks: N/A

B Use Or Create Scientific Artifacts: Yes

B1 Cite Creators Of Artifacts: Yes

B1 Elaboration: Sections 3.1 and 3.3

B2 Discuss The License For Artifacts: N/A

B3 Artifact Use Consistent With Intended Use: N/A

B4 Data Contains Personally Identifying Info Or Offensive Content: N/A

B5 Documentation Of Artifacts: N/A

B6 Statistics For Data: Yes

B6 Elaboration: Appendix A

C Computational Experiments: Yes

C1 Model Size And Budget: Yes

C1 Elaboration: Appendix B

C2 Experimental Setup And Hyperparameters: Yes

C2 Elaboration: Appendix B

C3 Descriptive Statistics: Yes

C3 Elaboration: Section 4

C4 Parameters For Packages: N/A

D Human Subjects Including Annotators: No

D1 Instructions Given To Participants: N/A

D2 Recruitment And Payment: N/A

D3 Data Consent: N/A

D4 Ethics Review Board Approval: N/A

D5 Characteristics Of Annotators: N/A

E Ai Assistants In Research Or Writing: No

E1 Information About Use Of Ai Assistants: N/A

Author Submission Checklist: yes

Submission Number: 603