Corpus for Modeling User Interactions in Online Persuasive Discussions

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 1135–1141 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 1135 Corpus for Modeling User Interactions in Online Persuasive Discussions Ryo Egawa, Gaku Morio, Katsuhide Fujita Tokyo University of Agriculture and Technology Koganei, Tokyo, Japan egawa@katfuji.lab.tuat.ac.jp, gakumorio@gmail.com, katfuji@cc.tuat.ac.jp Abstract Persuasions are common in online arguments such as discussion forums. To analyze persuasive strategies, it is important to understand how individuals construct posts and comments based on the semantics of the argumentative components. In addition to understanding how we construct arguments, understanding how a user post interacts with other posts (i.e., argumentative inter-post relation) still remains a challenge. Therefore, in this study, we developed a novel annotation scheme and corpus that capture both user-generated inner-post arguments and inter-post relations between users in ChangeMyView, a persuasive forum. Our corpus consists of arguments with 4612 elementary units (EUs) (i.e., propositions), 2713 EU-to-EU argumentative relations, and 605 inter-post argumentative relations in 115 threads. We analyzed the annotated corpus to identify the characteristics of online persuasive arguments, and the results revealed persuasive documents have more claims than non-persuasive ones and different interaction patterns among persuasive and non-persuasive documents. Our corpus can be used as a resource for analyzing persuasiveness and training an argument min- ing system to identify and extract argument structures. The annotated corpus and annotation guidelines have been made publicly available. Keywords: Argument Mining, Argument Schemes, ChangeMyView 1. Introduction In recent years, there have been advances in argument min- ing, a research field that analyzes argumentation or argu- mentative structures in a text using natural language pro- cessing. Several studies have proposed argument min- ing schemes (Stab and Gurevych, 2014; Habernal and Gurevych, 2016b), providing datasets with human anno- tations. The majority of the studies focused mainly on the identification and classification of argumentative compo- nents (e.g., claim and premise) and the argumentative rela- tions between the components (e.g., support and attack) in argumentative documents (Palau and Moens, 2009; Walton, 2012; Stab and Gurevych, 2017; Habernal and Gurevych, 2017; Boltužić and Šnajder, 2016). Several recent studies have also focused on the analysis of argumentation in persuasions or debates (Habernal and Gurevych, 2016b; Tan et al., 2016; Habernal and Gurevych, 2016a; Persing and Ng, 2017; Musi and Aakhus, 2018; Hidey and McKeown, 2018; Ji et al., 2018; Durmus and Cardie, 2019; Gleize et al., 2019; Hidey et al., 2017). These studies aim to reveal how individuals provide arguments to persuade an opponent. In addition, they aim to predict an opponent’s strategy and identify the attributes that make arguments convincing. For example, Tan et al. (2016) proposed an epoch-making task to predict which arguments were more persuasive in an online persuasive forum. Hidey et al. (2017) provided a novel corpus on the Tan et al. (2016) to reveal persuasive interactions. However, with the exception of some studies, such as (Chakrabarty et al., 2019; Ghosh et al., 2014), there are insufficient resources and annotation schemes for analyzing persuasive interactions. …

Content truncated to 500 words