Syntopical Graphs for Computational Argumentation Tasks

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 1583–1595 August 1–6, 2021. ©2021 Association for Computational Linguistics 1583 Syntopical Graphs for Computational Argumentation Tasks Joe Barrow∗ University of Maryland jdbarrow@cs.umd.edu Rajiv Jain Adobe Research rajijain@adobe.com Nedim Lipka Adobe Research lipka@adobe.com Franck Dernoncourt Adobe Research dernonco@adobe.com Vlad I. Morariu Adobe Research morariu@adobe.com Varun Manjunatha Adobe Research vmanjuna@adobe.com Douglas W. Oard University of Maryland oard@umd.edu Philip Resnik University of Maryland resnik@umd.edu Henning Wachsmuth Paderborn University henningw@upb.de Abstract Approaches to computational argumentation tasks such as stance detection and aspect de- tection have largely focused on the text of in- dividual claims, losing out on potentially valu- able context from the broader collection of text. We present a general approach to these tasks motivated by syntopical reading, a reading pro- cess that emphasizes comparing and contrast- ing viewpoints in order to improve topic under- standing. To capture collection-level context, we introduce the syntopical graph, a data struc- ture for linking claims within a collection. A syntopical graph is a typed multi-graph where nodes represent claims and edges represent dif- ferent possible pairwise relationships, such as entailment, paraphrase, or support. Experi- ments applying syntopical graphs to stance de- tection and aspect detection demonstrate state- of-the-art performance in each domain, signif- icantly outperforming approaches that do not utilize collection-level information. 1 Introduction Collections of text about the same topic such as news articles and research reports often present a variety of viewpoints. Adler and Van Doren (1940) proposed a formalized manual process for under- standing a topic based on multiple viewpoints in their book, How to Read a Book, applying dialec- tics to collection browsing. This process consists of four levels of reading, the highest of which is syntopical reading. Syntopical reading is focused on understanding a core concept by reading a col- lection of works. It requires finding passages on the ∗⋆ Work done while interning at Adobe Research. core concept that agree or disagree with each other, defining the issues, and analyzing the discussion to gain a better understanding of the core concept. The goal of the paper at hand is to operationalize the syntopical reading process computationally in order to help individuals make sense of a collection of documents for a given topic. Viewed through the lens of computational argu- mentation, these documents state claims or con- clusions that can be grouped by the aspects of the topic they discuss as well as by the stance they con- vey towards the topic (Stede and Schneider, 2018). An individual aiming to form a thorough under- standing of the topic needs to get an overview of these viewpoints and their interactions. This may be hard even if adequate tool support for brows- ing the collection is available (Wachsmuth et al., 2017a; Stab et al., 2018; Chen et al., 2019). We seek to enable systems that are capable of reconstructing viewpoints within a collection, where a viewpoint is expressed as a …

Content truncated to 500 words