Context-aware Argumentative Relation Mining

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 1127–1137, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics Context-aware Argumentative Relation Mining Huy V. Nguyen Computer Science Department University of Pittsburgh Pittsburgh, PA 15260 hvn3@pitt.edu Diane J. Litman Computer Science Department and Learning Research and Development Center University of Pittsburgh Pittsburgh, PA 15260 dlitman@pitt.edu Abstract Context is crucial for identifying argumen- tative relations in text, but many argument mining methods make little use of contex- tual features. This paper presents context- aware argumentative relation mining that uses features extracted from writing top- ics as well as from windows of context sentences. Experiments on student essays demonstrate that the proposed features im- prove predictive performance in two argu- mentative relation classification tasks. 1 Introduction By supporting tasks such as automatically iden- tifying argument components1 (e.g., premises, claims) in text, and the argumentative relations (e.g., support, attack) between components, ar- gument (argumentation) mining has been studied for applications in different research fields such as document summarization (Teufel and Moens, 2002), opinion mining (Boltuˇzi´c and ˇSnajder, 2014), automated essay evaluation (Burstein et al., 2003), legal information systems (Palau and Moens, 2009), and policy modeling platforms (Florou et al., 2013). Given a pair of argument components with one component as the source and the other as the tar- get, argumentative relation mining involves deter- mining whether a relation holds from the source to the target, and classifying the argumentative func- tion of the relation (e.g., support vs. attack). Ar- 1There is no consensus yet on an annotation scheme for argument components, or on the minimal textual units to be annotated. We follow Peldszus and Stede (2013) and con- sider “argument mining as the automatic discovery of an ar- gumentative text portion, and the identification of the relevant components of the argument presented there.” We also bor- row their term “argumentative discourse unit” to refer to the textual units (e.g., text segment, sentence, clause) which are considered as argument components. Essay 73. Topic: Is image more powerful than the written word? ...(1)Hence, I agree only to certain de- gree that in today’s world, image serves as a more effective means of communica- tion[MajorClaim]. (2)Firstly, pictures can influence the way people think[Claim]. (3)For example, nowa- days horrendous images are displayed on the cigarette boxes to illustrate the con- sequences of smoking[Premise]. (4)As a result, statistics show a slight reduction in the number of smokers, indicating that they realize the effects of the negative habit[Premise]... Figure 1: Excerpt from a student persuasive essay (Stab and Gurevych, 2014a). Sentences are num- bered and argument components are tagged. gumentative relation mining - beyond argument component mining - is perceived as an essential step towards more fully identifying the argumenta- tive structure of a text (Peldszus and Stede, 2013; Sergeant, 2013; Stab et al., 2014). Consider the second paragraph shown in Figure 1. Only detect- ing the argument components (a claim in sentence 2 and two premises in sentences 3 and 4) does not …

Content truncated to 500 words