ENGLAWI: From Human- to Machine-Readable Wiktionary

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3016–3026 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3016 ENGLAWI: From Human- to Machine-Readable Wiktionary Franck Sajous, Basilio Calderone and Nabil Hathout CLLE-ERSS, CNRS & Université de Toulouse 2 {franck.sajous, basilio.calderone, nabil.hathout}@univ-tlse2.fr Abstract This paper introduces ENGLAWI, a large, versatile, XML-encoded machine-readable dictionary extracted from Wiktionary. ENGLAWI contains 752,769 articles encoding the full body of information included in Wiktionary: simple words, compounds and multiword expressions, lemmas and inflectional paradigms, etymologies, phonemic transcriptions in IPA, definition glosses and usage examples, translations, semantic and morphological relations, spelling variants, etc. It is fully documented, released under a free license and supplied with G-PeTo, a series of scripts allowing easy information extraction from ENGLAWI. Additional resources extracted from ENGLAWI, such as an inflectional lexicon, a lexicon of diatopic variants and the inclusion dates of headwords in Wiktionary’s nomenclature are also provided. The paper describes the content of the resource and illustrates how it can be – and has been – used in previous studies. We finally introduce an ongoing work that computes lexicographic word embeddings from ENGLAWI’s definitions. Keywords: Wiktionary, Machine-Readable Dictionary of Contemporary English, Definitions-based Embeddings 1. Introduction This paper introduces ENGLAWI, a structured and nor- malized version of the English Wiktionary encoded into a workable XML format. ENGLAWI is freely available for download,1 fully documented, and is supplied with G-PeTo (GLAWI Perl Tools), a set of scripts that helps extract spe- cific information from GLAWI dictionaries. Contrary to other approaches designed to extract particular data from Wiktionary, ENGLAWI is part of a series of works aim- ing to provide the full body of information encoded in the collaborative dictionaries. In previous papers, we pre- sented GLAWI (Sajous and Hathout, 2015; Hathout and Sajous, 2016), extracted from the French Wiktionnaire and GLAWIT (Calderone et al., 2016), extracted from the Ita- lian Wikizionario. In these papers, we described the con- version and the standardization of the heterogeneous data extracted from the dictionaries’ dumps and we illustrated how specific lexicons can be easily tailored on demand. We did adopt an identical approach to develop ENGLAWI. The current paper describes the content of ENGLAWI, il- lustrates how it is structured and suggests possible uses for linguistic studies and NLP applications. The remainder of the article is organized as follows. In Section 2, we discuss the usefulness of machine-readable dictionaries (MRD) for linguistics and NLP. We compare in Section 3 our approach to related works on information extraction from Wiktionary. Then, we describe the struc- ture of ENGLAWI in Section 4 and we introduce additional tools and resources provided with ENGLAWI in Section 5. Finally, we recap in Section 6 how ENGLAWI, GLAWI and GLAWIT have been used in various linguistic and NLP works and we propose a new method for computing lexico- graphic word embeddings from ENGLAWI’s definitions. 2. Are MRD of any use for NLP? In a six-week online course on English dictionaries running in September …

Content truncated to 500 words