DiMLex-Bangla: A Lexicon of Bangla Discourse Connectives

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 1097–1102 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 1097 DiMLex-Bangla: A Lexicon of Bangla Discourse Connectives Debopam Das, Manfred Stede, Soumya Sankar Ghosh, Lahari Chatterjee Humboldt University of Berlin, University of Potsdam, Mihup Communications, Jadavpur University dasdebop@bu-berlin.de, stede@uni-potsdam.de, soumya@mihup.com, lahari.chatterjee@gmail.com Abstract We present DiMLex-Bangla, a newly developed lexicon of discourse connectives in Bangla. The lexicon, upon completion of its first version, contains 123 Bangla connective entries, which are primarily compiled from the linguistic literature and translation of English discourse connectives. The lexicon compilation is later augmented by adding more connectives from a currently developed corpus, called the Bangla RST Discourse Treebank (Das and Stede, 2018). DiMLex-Bangla provides information on syntactic categories of Bangla connectives, their discourse semantics and non-connective uses (if any). It uses the format of the German connective lexicon DiMLex (Stede and Umbach, 1998), which provides a cross-linguistically applicable XML schema. The resource is the first of its kind in Bangla, and is freely available for use in studies on discourse structure and computational applications. Keywords: discourse connectives, lexicon, discourse relation, Bangla 1. Introduction Successful interpretation of texts (at the cognitive level) and the establishment of valid text structures (at the meta- discourse level) largely depend on the identification of the presence and type of discourse relations that hold between text segments. Discourse relations are often indicated by discourse connectives (and, but, however, thus, etc.), which are widely believed to be the most explicit, most prototypical and most reliable relational signals (Danlos et al., 2018). Accordingly, the role of connectives in discourse interpretation has been extensively studied in both (psycholinguistic) discourse processing (see Kleijn et al. (2019) for an overview) and computational discourse applications such as discourse parsing (Hernault et al., 2010; Lin et al., 2014), machine translation (Meyer et al., 2011), text summarization (Alemany, 2005), or argumen- tation mining (Kirschner et al., 2015). This has led to the production of an increasingly large number of text corpora in different languages, annotated for discourse connectives (most often, in addition to other signals), based on different discourse frameworks (for references, see Das and Taboada (2019)). Parallel to the practice of developing discourse-annotated corpora, initiatives have also been taken to construct ancil- lary resources such as lexicons of discourse connectives, which can provide the list of connectives in a language, along with useful information about their syntactic and semantic-pragmatic properties. Such lexicons are devel- oped and are becoming increasingly available in different languages, beginning with DiMLex in German (Stede and Umbach, 1998), later DPDE for Spanish (Briz et al., 2008) and LexConn for French (Roze et al., 2012), and more recently LICO for Italian (Feltracco et al., 2016), CzeDLex for Czech (M´ırovsk´y et al., 2017), LDM-PT for Portuguese (Mendes et al., 2018), DisCoDict for Dutch (Bourgonje et al., 2018) and DiMLex-Eng for English (Das et al., 2018). We extend this line of work by developing a new lexicon of discourse connectives …

Content truncated to 500 words