Overview of the CAIL 2023 Argument Mining Track

TL;DR

This paper surveys the CAIL 2023 Argument Mining Track, a two-stage task that identifies and extracts interacting plaintiff–defense argument pairs in Chinese trial dialogs. It introduces the CAIL2023-Argmine dataset, expanded from SMP-CAIL2020-Argmine with nine new causes of action and multimodal data, and presents evaluation protocols and baseline results. The top submissions largely rely on pretrained language models with domain adaptation, data augmentation, adversarial training, and ensembling, achieving clear gains over the baseline, though Stage 2 remains more challenging and the multimodal bonus data saw limited use. The study provides actionable insights into task formulation, model design, and data practices for judicial argument mining, while outlining directions for future work in multimodal judicial understanding and broader cause-action coverage.

Abstract

We give a detailed overview of the CAIL 2023 Argument Mining Track, one of the Chinese AI and Law Challenge (CAIL) 2023 tracks. The main goal of the track is to identify and extract interacting argument pairs in trial dialogs. It mainly uses summarized judgment documents but can also refer to trial recordings. The track consists of two stages, and we introduce the tasks designed for each stage; we also extend the data from previous events into a new dataset -- CAIL2023-ArgMine -- with annotated new cases from various causes of action. We outline several submissions that achieve the best results, including their methods for different stages. While all submissions rely on language models, they have incorporated strategies that may benefit future work in this field.

Figures (1)

• Figure 1: Score distribution of teams with valid submissions for the second stage.