SAD: A Large-Scale Strategic Argumentative Dialogue Dataset

TL;DR

This paper introduces SAD, a large-scale dataset of multi-turn strategic argumentative dialogues, enabling models to generate contextually appropriate arguments with diverse strategies, advancing research in argumentation dialogue modeling.

Contribution

The paper presents the first large-scale dataset of strategic argumentative dialogues with annotated strategies, supporting multi-turn, context-aware argument generation.

Findings

Pretrained models can generate contextually relevant arguments.

Analysis reveals diverse strategic usage patterns.

SAD dataset enables deeper modeling of argumentative dialogue.

Abstract

Argumentation generation has attracted substantial research interest due to its central role in human reasoning and decision-making. However, most existing argumentative corpora focus on non-interactive, single-turn settings, either generating arguments from a given topic or refuting an existing argument. In practice, however, argumentation is often realized as multi-turn dialogue, where speakers defend their stances and employ diverse argumentative strategies to strengthen persuasiveness. To support deeper modeling of argumentation dialogue, we present the first large-scale \textbf{S}trategic \textbf{A}rgumentative \textbf{D}ialogue dataset, SAD, consisting of 392,822 examples. Grounded in argumentation theories, we annotate each utterance with five strategy types, allowing multiple strategies per utterance. Unlike prior datasets, SAD requires models to generate contextually appropriate arguments conditioned on the dialogue history, a specified stance on the topic, and targeted argumentation strategies. We further benchmark a range of pretrained generative models on SAD and present in-depth analysis of strategy usage patterns in argumentation.