Some Issues with Building a Multilingual Wordnet

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3189–3197 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3189 Some Issues with Building a Multilingual Wordnet Francis Bond♠, Luis Morgado da Costa♠, Michael Wayne Goodman♠, John P. McCrae♢, Ahti Lohk♣ ♠ Nanyang Technological University (NTU), Singapore ♢ Data Science Institute, National University of Ireland Galway, ♣ Tallinn University of Technology, Estonia bond@ieee.org Abstract In this paper we discuss the experience of bringing together over 40 different wordnets. We introduce some extensions to the GWA wordnet LMF format proposed in Vossen et al. (2016) and look at how this new information can be displayed. Notable extensions include: confidence, corpus frequency, orthographic variants, lexicalized and non-lexicalized synsets and lemmas, new parts of speech, and more. Many of these extensions already exist in multiple wordnets – the challenge was to find a compatible representation. To this end, we introduce a new version of the Open Multilingual Wordnet (Bond and Foster, 2013), that integrates a new set of tools that tests the extensions introduced by this new format, while also ensuring the integrity of the Collaborative Interlingual Index (CILI: Bond et al., 2016), avoiding the same new concept to be introduced through multiple projects. Keywords: multilingual lexicon, wordnet, collaborative development 1. Introduction This paper provides a summary and update of some of the issues involved with coordinating multiple lexical wordnets. The Princeton WordNet (PWN) is one of the most cited lex- ical resources in the world with over 1.8 as many citations as the most cited paper in the ACL anthology1 (Marcus et al. (2004) according to Joseph and Radev (2007)). This suc- cess has inspired wordnets in many languages, and many attempts to link them, such as EuroWordnet (Vossen, 1998), the Asian Language Wordnet (Charoenporn et al., 2008) and many more. In this paper, we discuss the approach of the Open Multilingual Wordnet (Bond and Foster, 2013). In the first version of this project (OMW 1.0), wordnets were linked through the PWN. Although we knew the approach to be fundamentally incorrect (different languages lexical- ize different concepts and link them in different ways (Fell- baum and Vossen, 2012)), it was chosen as a first approx- imation for multiple reasons. The main one was that most existing wordnets are built by translating the PWN: the ex- tend model (Vossen, 1998). For example, dogn:1 is linked to the lemmas chien in French, anjing in Malay, and so on. The overall structure of PWN serves as a useful scaffold and the fact that, for example, a dogn:1 is an animaln:1 is language independent. The main innovations of the OMW 1.0 were an emphasis on open licenses (so that all the data could be legally shared) and a simple shared format (so that resources could be easily converted). The second version of the OMW (2.0) revived the idea of the InterLingual Index (ILI: Vossen et al., 1999) with the Col- laborative Interlingual Index (CILI: Bond et al., 2016). …

Content truncated to 500 words