Towards automatic enrichment of standardized electronic dictionaries by semantic classes

The 2014 Conference on Computational Linguistics and Speech Processing ROCLING 2014, pp. 96-109  The Association for Computational Linguistics and Chinese Language Processing Towards automatic enrichment of standardized electronic dictionaries by semantic classes Elleuch Imen, Gargouri Bilel and Ben Hamadou Abdelmajid1 Abstract In this paper we propose an approach for the automatic enrichment of standardized electronic dictionaries by the semantic classes. This approach consists of three phases. The first phase treat the semantic classification process founded on the studies of Gaston Gross. The second phase profites from the existed subject fields in the dictionary's lexical entries in order to attribute the suitable semantic classes. The final phase realizes syntactic analyses of the textual content of meanings’s lexical entries. This phase, aims to refine the subject field based enrichment and also treats the non enriched meanings in the second phase. In addition, it attributes the same semantic classes for the synonym meanings. We used an available standardized Arabic dictionary to tested the performance of the proposed approach. Keywords: Automatic enrichment, standardized electronic dictionnaries, semantic classes, Arabic language. 1. Introduction Semantic knowledge, especially semantic classes which aim to characterize meanings of lexical units in dictionaries, have attracted considerable interest in both linguistic (Stede, 1998), (Dorr, 1997) and computational linguistics (Kipper et al 2000). Such semantic class can be definite as a semantic linguistic propriety classifying meanings and can therefore be used as a valuable means of comprehending the specific meaning of polysimous lexical units. Thus the need of dictionaries with semantic classes has become a necessity for Natural Language Processing (NLP) applications. For various languages, various semantic classifications are now available. We can list the verbs classification (Pinker, 1989; Jackendoff, 1990; Levin, 1993, Dubois and Dubois-Charlier, 1997) that regroups together verbs that share both a common semantics and a set of syntactic alternations. Also, we notice WordNet (Fellbaum, 1998) that provides semantic ontological classification and FrameNet (Fillmore, 1985) that hierarchically classify lexical units using various relationship as synonymy, antonym and is-a relations. However, the referential classification is based on semantic features like [+/- human], [+/- concrete], etc. characterizing semantically each lexical unit outside of the meaning’s contexts. Object classes (Gross, 1994) defines a semantic classification based on surface realization of predicate argument structure. A semantic class groups together predicates as arguments having the same syntactic constrictions. Rely on a semantic classification; two methods of enrichment lexical resources by semantic classes exist. The first one is manual. It is characterized by the large number of lexical units to be classified FrameNet  FSEGS, B.P. 1088, 3018 Sfax, Tunisia E-mail: { imen.elleuch, bilel.gargouri}@fsegs.rnu.tn  ISIMS, B.P. 242, 3021 Sakiet-Ezzit Sfax, Tunisia E-mail: abdelmajid.benhamadou@isimsf.rnu.tn 96 (Fillmore, 1985) this is why it is a costly and time-consuming method. The second method is automatic. It can use corpora (Fuchs & Habert, 2004), (Condamines, 2005) or in some cases, texts of the treated lexical resources (Rastier, 2001) and (Valette et al, 2006). The automatic method does not necessitate the intervention of the human expert during the enrichment process …

Content truncated to 500 words