KNOWCOMP POKEMON Team at DialAM-2024: A Two-Stage Pipeline for Detecting Relations in Dialogical Argument Mining
Zihao Zheng, Zhaowei Wang, Qing Zong, Yangqiu Song
TL;DR
This paper addresses dialogical argument mining by introducing a two-stage pipeline that first detects propositional relations among I-nodes via a Two-Step S-node Prediction Model and then identifies illocutionary relations via a YA-node Prediction Model with contextual information. Data augmentation is employed to balance relation and non-relation examples, and experiments compare multiple transformer baselines, showing the two-stage approach improves generalization. The proposed method achieves top ARI Focused scores and competitive Global Focused scores on the DialAM-2024 task, demonstrating the benefit of context and staged reasoning in complex dialogue-structured arguments. The work highlights practical gains for extracting structured argumentative relations from dialogic content and suggests future work in incorporating richer locution features and broader model families.
Abstract
Dialogical Argument Mining(DialAM) is an important branch of Argument Mining(AM). DialAM-2024 is a shared task focusing on dialogical argument mining, which requires us to identify argumentative relations and illocutionary relations among proposition nodes and locution nodes. To accomplish this, we propose a two-stage pipeline, which includes the Two-Step S-Node Prediction Model in Stage 1 and the YA-Node Prediction Model in Stage 2. We also augment the training data in both stages and introduce context in Stage 2. We successfully completed the task and achieved good results. Our team Pokemon ranked 1st in the ARI Focused score and 4th in the Global Focused score.