A Dataset of Translational Equivalents Built on the Basis of plWordNet-Princeton WordNet Synset Mapping

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3260–3264 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3260 A Dataset of Translational Equivalents Built on the Basis of plWordNet-Princeton WordNet Synset Mapping Ewa Rudnicka, Tomasz Naskr˛et Wroclaw University of Science and Technology, Faculty of Computer Science and Management Wybrzeże Wyspiańskiego 27, 50-370 Wroclaw {ewa.rudnicka, tomasz.naskret}@pwr.edu.pl Abstract The paper presents a dataset of 11,000 Polish-English translational equivalents in the form of pairs of plWordNet and Princeton WordNet lexical units (senses) linked by three types of equivalence links: strong equivalence, regular equivalence, and weak equivalence. The resource consists of the two subsets. The first subset was built in result of manual annotation of an extended sample of Polish-English sense pairs partly randomly extracted from synsets linked by interlingual relations such as I-synononymy, I-partial synonymy and I-hyponymy and partly manually selected from the surrounding synsets in the hypernymy hierarchy. The second subset was created as a result of the manual checkup of an automatically generated lists of pairs of sense equivalents on the basis of a couple of simple, rule-based heuristics. For both subsets, the same methodology of equivalence annotation was adopted based on the verification of a set of formal, semantic-pragmatic and translational features. The constructed dataset is a novum in the wordnet domain and can facilitate the precision of bilingual NLP tasks such as automatic translation, bilingual word sense disambiguation and sentiment annotation. Keywords: wordnet, bilingual alignment, senses 1. Introduction Bi- and multilingual wordnets are used in a variety of tasks, such as automatic and manual translation, bilingual senti- ment annotation and word sense disambiguation (Google Translate, (Bond et al., 2019)). More and more people also reach to them in addition to, or instead of, electronic dictionaries. Wordnets usually provide pairs or groups of conceptual equivalents based on mapping between synsets (sets of synonymous lexical units, also called senses). Still, such mapping may be of different granularity depending on the granularity of synsets involved (that is the number of lex- ical units (senses) per synset). It may include 1-1 mappings (for singleton synsets), 1-to-many mappings (for singleton and multi-unit synsets) and many-to-many mappings (for multi-unit synsets). Intuitively, for all the tasks mentioned, we predict that precision should rise directly proportional to the precision of bilingual links. The most precise links ob- viously hold between singleton synsets. In fact, this moves mapping from the level of synsets to the level of individ- ual sense pairs. However, such mappings are rather rare in the wordnet domain, because its automatization is difficult, while manual work extremely costly and time-consuming. The notable exceptions are the works of (Copestake et al., 1994) and our own (Rudnicka et al., 2017b), (Rudnicka et al., 2017a), (Rudnicka et al., 2018). In our most recent paper (Rudnicka et al., 2019), we have re- fined our earlier feature-based typology of equivalence types together with the method of their application and have tested it on a sample of lexical unit pairs. In …

Content truncated to 500 words