Visser Jacky, Lawrence John, Reed Chris, Wagemans Jean, Walton Douglas
Centre for Argument Technology, University of Dundee, Nethergate, Dundee, DD1 4HN UK.
Amsterdam Centre for Language and Communication (ACLC), University of Amsterdam, Spuistraat 134, Amsterdam, 1012 VB The Netherlands.
Argumentation. 2021;35(1):101-139. doi: 10.1007/s10503-020-09519-x. Epub 2020 May 7.
Argument schemes are abstractions substantiating the inferential connection between premise(s) and conclusion in argumentative communication. Identifying such conventional patterns of reasoning is essential to the interpretation and evaluation of argumentation. Whether studying argumentation from a theory-driven or data-driven perspective, insight into the actual use of argumentation in communicative practice is essential. Large and reliably annotated corpora of argumentative discourse to quantitatively provide such insight are few and far between. This is all the more true for argument scheme corpora, which tend to suffer from a combination of limited size, poor validation, and the use of ad hoc restricted typologies. In the current paper, we describe the annotation of schemes on the basis of two distinct classifications: Walton's taxonomy of argument schemes, and Wagemans' Periodic Table of Arguments. We describe the annotation procedure for each, and the quantitative characteristics of the resulting annotated text corpora. In doing so, we extend the annotation of the preexisting US2016 corpus of televised election debates, resulting in, to the best of our knowledge, the two largest consistently annotated corpora of schemes in argumentative dialogue publicly available. Based on evaluation in terms of inter-annotator agreement, we propose further improvements to the guidelines for annotating schemes: the argument scheme key, and the Argument Type Identification Procedure.
论证型式是对论证性交流中前提与结论之间推理联系进行实质化的抽象形式。识别这种常规推理模式对于论证的解释和评估至关重要。无论从理论驱动还是数据驱动的角度研究论证，洞察论证在交际实践中的实际运用都至关重要。能定量提供此类洞察的大规模且标注可靠的论证性话语语料库少之又少。对于论证型式语料库而言更是如此，这类语料库往往存在规模有限、验证不足以及使用特设的受限类型学等问题。在本文中，我们基于两种不同的分类描述了型式的标注：沃尔顿的论证型式分类法和瓦格曼斯的论证周期表。我们描述了每种分类的标注程序以及由此产生的标注文本语料库的量化特征。在此过程中，我们扩展了对已有的2016年美国电视选举辩论语料库的标注，据我们所知，由此产生了两个公开可用的、规模最大且标注一致的论证性对话型式语料库。基于对标注者间一致性的评估，我们对型式标注指南提出了进一步改进建议：论证型式键和论证类型识别程序。
