Discovering Fuzzy Synsets from the Redundancy in Different Lexical-Semantic Resources

Discovering Fuzzy Synsets from the Redundancy in Different Lexical-Semantic Resources Hugo Gonc¸alo Oliveira, F´abio Santos CISUC, Department of Informatics Engineering University of Coimbra, Portugal hroliv@dei.uc.pt, fasantos@student.dei.uc.pt Abstract Although represented as such in wordnets, word senses are not discrete. To handle word senses as fuzzy objects, we exploit the graph structure of synonymy pairs acquired from different sources to discover synsets where words have different membership degrees that reflect confidence. Following this approach, a wide-coverage fuzzy thesaurus was discovered from a synonymy network compiled from seven Portuguese lexical-semantic resources. Based on a crowdsourcing evaluation, we can say that the quality of the obtained synsets is far from perfect but, as expected in a confidence measure, it increases significantly for higher cut-points on the membership and, at a certain point, reaches 100% correction rate. Keywords: wordnets, synonyms, word senses, fuzzy synsets, clusters 1. Introduction Wordnets are lexical-semantic knowledge bases, modelled after Princeton WordNet (Fellbaum, 1998), where words are grouped with their synonyms, in the so-called synsets, a representation of natural language concepts by their pos- sible lexicalisations. As natural language is ambiguous, dif- ferent words might have the same meaning, and the same word might be in more than one synset, one for each of its senses. But word senses are not discrete (Kilgarriff, 1996). They are complex and overlapping structures and their representation as crisp objects does not reflect the human language. A more realistic approach would han- dle word senses with uncertainty and represent concepts as fuzzy synsets. In fact, this idea is not new. Fuzzy member- ships of words to synsets have been obtained from manual judgements (Borin and Forsberg, 2010) and there have been approaches for representing WordNet as an ontology with fuzzy synsets and relations (Ara´uz et al., 2012). The wordnet model has been adopted by many languages (see e.g. Bond and Paik (2012)) and there are some with more than one wordnet, including Portuguese, for which there are six wordnets, created by independent teams, following different approaches, and with different licenses (Gonc¸alo Oliveira et al., 2015). But, in opposi- tion to English, where Princeton WordNet can be seen as a standard resource, the open Portuguese wordnets are still in an early development stage and all of them have their strengths and limitations, either in terms of coverage or cor- rection. For instance, OpenWN-PT (de Paiva et al., 2012) and, especially, PULO (Sim˜oes and Guinovart, 2014), still have a low coverage of words, senses and relation types. They both have a controlled expansion and are aligned with Princeton WordNet where additional information can be obtained. On the other hand, Onto.PT (Gonc¸alo Oliveira and Gomes, 2014) is the largest, but has more reliability is- sues, because it is created automatically, from the exploita- tion of dictionaries and other textual sources. Its creation follows the ECO approach, where relations, synsets and their boundaries, as well as relation attachments, are dis- covered automatically in three steps: (i) extraction, where semantic relations of different types, held between words, are acquired from …

Content truncated to 500 words