DiscSense: Automated Semantic Analysis of Discourse Markers

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 991–999 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 991 DiscSense: Automated Semantic Analysis of Discourse Markers Damien Sileo3,1, Tim Van de Cruys2, Camille Pradel1, Philippe Muller3 1:Synapse D´eveloppement, 2:IRIT (CNRS), 3:IRIT (University of Toulouse) damien.sileo@irit.fr Abstract Discourse markers (by contrast, happily, etc.) are words or phrases that are used to signal semantic and/or pragmatic relationships between clauses or sentences. Recent work has fruitfully explored the prediction of discourse markers between sentence pairs in order to learn accurate sentence representations, that are useful in various classification tasks. In this work, we take another perspective: using a model trained to predict discourse markers between sentence pairs, we predict plausible markers between sentence pairs with a known semantic relation (provided by existing classification datasets). These predictions allow us to study the link between discourse markers and the semantic relations annotated in classification datasets. Handcrafted mappings have been proposed between markers and discourse relations on a limited set of markers and a limited set of categories, but there exist hundreds of discourse markers expressing a wide variety of relations, and there is no consensus on the taxonomy of relations between competing discourse theories (which are largely built in a top-down fashion). By using an automatic prediction method over existing semantically annotated datasets, we provide a bottom-up characterization of discourse markers in English. The resulting dataset, named DiscSense, is publicly available. Keywords: Discourse marker semantics, pragmatics, discourse marker prediction 1. Motivation Discourse markers are a common language device used to make explicit the semantic and/or pragmatic relationships between clauses or sentences. For example, the marker so in sentence (1) indicates that the second clause is a conse- quence of the first. (1) We’re standing in gasoline, so you should not smoke. Several resources enumerate discourse markers and their use in different languages, either in discourse marker lexi- cons (Knott, 1996; Stede, 2002; Roze et al., 2012; Das et al., 2018) or in corpora, annotated with discourse relations, such as the well-known English Penn Discourse TreeBank (Prasad et al., 2008), which inspired other efforts in Turk- ish, Chinese and French (Zeyrek and Webber, 2008; Zhou et al., 2014; Danlos et al., 2015). The PDTB identifies dif- ferent types of discourse relation categories (such as con- junction and contrast) and the respective markers that fre- quently instantiate these categories (such as and and how- ever, respectively), and organizes them in a three-level hi- erarchy. It must be noted, however, that there is no gen- eral consensus on the typology of these markers and their rhetorical functions. As such, theoretical alternatives to the PDTB exist, such as Rhetorical Structure Theory or RST (Carlson et al., 2001), and Segmented Discourse Repre- sentation Theory or SDRT (Asher and Lascarides, 2003). Moreover, marker inventories focus on a restricted number of rhetorical relations that are too coarse and not exhaus- tive, since discourse marker use depends on the grammat- ical, stylistic, pragmatic, semantic …

Content truncated to 500 words