Tools for Building an Interlinked Synonym Lexicon Network

Tools for Building an Interlinked Synonym Lexicon Network Zdeˇnka Urešová, Eva Fuˇcíková, Eva Hajiˇcová, Jan Hajiˇc Charles University, Faculty of Mathematics and Physics Institute of Formal and Applied Linguistics Malostranské nám. 25 11800 Prague 1, Czech Republic {uresova,fucikova,hajicova,hajic}@ufal.mff.cuni.cz Abstract This paper presents the structure, features and design of a new interlinked verbal synonym lexical resource called CzEngClass and the editor tool being developed to assist the work. This lexicon captures cross-lingual (Czech and English) synonyms, using valency behavior of synonymous verbs in relation to semantic roles as one of the criteria for defining such interlingual synonymy. The tool, called Synonym Class Editor - SynEd, is a user-friendly tool specifically customized to build and edit individual entries in the lexicon. It helps to keep the cross-lingual synonym classes consistent and linked to internal as well as to well-known external lexical resources. The structure of SynEd also allows to keep and edit the appropriate syntactic and semantic information for each Synonym Class member. The editor makes it possible to display examples of class members’ usage in translational context in a parallel corpus. SynEd is platform independent and may be used for multiple languages. SynEd, CzEngClass and services based on them will be openly available. Keywords: Lexical Resource, Parallel Corpus, Semantics, Syntax, Synonymy, Valency 1. Introduction We present a demonstration of our lexicon editor, called SynEd, for creating an interlinked multilingual (for the time being, bilingual) lexical resource—a contextually- based synonym lexicon of verbs based on their syntac- tic and semantic behavior in (bilingual) context. We also present the design and structure (scheme) of the lexicon dataset proper. The lexicon, under the working name “CzEngClass” (Ure- šová et al., 2018), is being built “bottom-up”, using a par- allel bilingual corpus with a rich manual annotation (the Prague Czech-English Dependency Treebank, version 2.0), its associated valency lexicons, i.e., EngVallex, PDT-Vallex and CzEngVallex, and it is being linked to other exter- nal lexical resources–VALLEX, FrameNet, VerbNet, Prop- Bank and Czech and English WordNets (Sect. 3.). One of the crucial parts of the “CzEngClass” project was to establish a scheme for this resource (Sect. 4.), taking into account not only its intended multilinguality, but also links to the initial bilingual parallel corpus (and more corpora in the future) and the existing relevant lexical resources. The semantic role labels used for defining the individual syn- onym classes are based mostly on FrameNet (and in some cases on VerbNet), while the argument labels and their mor- phosyntactic behavior are taken from the valency lexicons used for the PCEDT annotation. One of the main princi- ples for the design of “CzEngClass” follows A. Kilgariff’s idea of corpus - dictionary linkage (cf. the PDIC and PCID model described in (Kilgariff, 2005)), so we are strictly keeping references to all of the used lexical resources (in- ternal and external) as well as to the corpus examples influ- encing the class divisions. Since the goal of the project is to identify contextually- based Czech and English synonyms, each verb has to be …

Content truncated to 500 words