PortiLexicon-UD: a Portuguese Lexical Resource according to Universal Dependencies Model

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6635–6643 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6635 PortiLexicon-UD: a Portuguese Lexical Resource according to Universal Dependencies Model Lucelene Lopes1, Magali S. Duran1, Paulo Fernandes2, Thiago A. S. Pardo1 1Interinstitutional Center for Computational Linguistics (NILC) Institute of Mathematical and Computer Sciences, University of S˜ao Paulo, S˜ao Carlos–SP, Brazil 2Merrimack College, North Andover–MA, USA {lucelene, magali.duran, lemelle.fernandes}@gmail.com, taspardo@icmc.usp.br Abstract This paper presents PortiLexicon-UD, a large and freely available lexicon for Portuguese delivering morphosyntactic information according to the Universal Dependencies model. This lexical resource includes part of speech tags, lemmas, and morphological information for words, with 1,221,218 entries (considering word duplication due to different combination of PoS tag, lemma, and morphological features). We report the lexicon creation process, its computational data structure, and its evaluation over an annotated corpus, showing that it has a high language coverage and good quality data. Keywords: Lexical Resources, Portuguese, Universal Dependencies 1. Introduction This paper describes the construction of a robust lex- ical resource for Brazilian Portuguese – PortiLexicon- UD – with 1,221,218 distinct entries in the form of 4- tuples, composed by information on word, lemma, Part of Speech (PoS) tag and morphological features, fol- lowing the Universal Dependencies (UD) model and guidelines (Nivre et al., 2020). Lexical resources are useful for several different tasks in Natural Language Processing (NLP), from support- ing PoS tagging and parsing to named entity recog- nition and information extraction, among many other possibilities. They may also subsidize linguistic studies and corpus linguistics initiatives, helping characteriz- ing specific text genres and domains, and allowing the investigation of patterns of word usage. In particular, the lexicon we propose also contributes to UD-oriented research for Portuguese, in line with what has been done for other languages (Przepi´orkowski and Patejuk, 2018; Cecchini et al., 2020; Miletic et al., 2020; Sed- dah et al., 2020). The new resource was based on the UNITEX-PB (Mu- niz, 2004) lexicon as a starting point, first by doing a PoS tagset mapping from more traditional tags to UD tags, and then by incorporating new decisions, which will be commented on later. To the best of our knowledge, there is no available lexical resource using UD model for Portuguese. In fact, Portuguese resources in UD are limited to an- notated corpora, three of which are publicly available at UD site: BOSQUE-UD (Rademaker et al., 2017), a subset of Floresta Sin(c)t´atica with 210,963 tokens, PUD (Uszkoreit et al., 2017), a small parallel corpus created for CoNLL 2017 shared task, with 21,917 to- kens, and GSD (McDonald et al., 2013), a Portuguese translation from Google UD treebank, with 297,057 to- kens. This paper goal is to present a lexical resource, called PortiLexicon-UD, which contains 858,922 unique Por- tuguese words (homographs are considered only once) and 1,221,218 entries considering word duplication due to different combinations of PoS tag, lemma and mor- phological features. The resource, available for online search and for download …

Content truncated to 500 words