Towards Better Non-Tree Argument Mining: Proposition-Level Biaffine Parsing with Task-Specific Parameterization

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3259–3266 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 3259 Towards Better Non-Tree Argument Mining: Proposition-Level Biaffine Parsing with Task-Specific Parameterization Gaku Morio, Hiroaki Ozaki, Terufumi Morishita, Yuta Koreeda and Kohsuke Yanai Hitachi, Ltd. Research and Development Group Kokubunji, Tokyo, Japan {gaku.morio.vn, hiroaki.ozaki.yu, terufumi.morishita.wp, yuta.koreeda.pb, kohsuke.yanai.cs}@hitachi.com Abstract State-of-the-art argument mining studies have advanced the techniques for predicting argu- ment structures. However, the technology for capturing non-tree-structured arguments is still in its infancy. In this paper, we focus on non-tree argument mining with a neural model. We jointly predict proposition types and edges between propositions. Our proposed model in- corporates (i) task-specific parameterization (TSP) that effectively encodes a sequence of propositions and (ii) a proposition-level bi- affine attention (PLBA) that can predict a non-tree argument consisting of edges. Exper- imental results show that both TSP and PLBA boost edge prediction performance compared to baselines. 1 Introduction Argument mining, a research area that focuses on predicting argumentation structures in a text, has been receiving much attention. To date, efforts in argument mining were devoted to predicting tree arguments in which a claim proposition is repre- sented as a root and premise propositions are repre- sented as leaves. For example, Stab and Gurevych (2017) introduced Argument Annotated Essays (hereafter, Essay), and researchers attempted to predict tree arguments in the corpus (Eger et al., 2017; Potash et al., 2017; Kuribayashi et al., 2019). However, these techniques lack the capability of dealing with more flexible arguments such as reason edges where a proposition can have several parents. To this end, Park and Cardie (2018) pro- vided a less restrictive argument mining dataset known as Cornell eRulemaking Corpus (CDCP), which contains flexible edges (see VALUES (a), (b), and TESTIMONY (e) in Figure 1). Figure 2 shows a distribution of outgoing edges for Essay and CDCP. Propositions in CDCP have sparse con- nections, making the majority of propositions iso- ... [ I'm with Massachusetts on this one. ]a ... [ Repetitive and robo - calls are annoying and not productive. ]b ... [ Another fact about robo - calls is that their messages often start in the middle, ]c ... [ or maybe this is done on purpose. ]d ... [ When it has happened to me, I just hang up. ]e ... [ Policies regulating the number of contacts made within a specific time period should include all modes of technology. ]f VALUE�(a) VALUE�(b) FACT�(c) VALUE�(d) TESTIMONY�(e) POLICY�(f) REASON REASON REASON Figure 1: Example graph in the CDCP corpus 0 1 2 3 4 5 #outgoing edge 0 2000 4000 frequency Essay CDCP Figure 2: Distribution of the outgoing edges (i.e., Sup- port/Attack or REASON/EVIDENCE relations) from a node (proposition) in Essay and CDCP corpora lated from the others. Besides, a proposition in Essay has at most one outgoing edge, while that in CDCP has a variable number of edges (i.e., there are about 200 propositions which have two or more …

Content truncated to 500 words