Towards detecting anomalies in the content of standardized LMF dictionaries

Proceedings of Recent Advances in Natural Language Processing, pages 719–726, Hissar, Bulgaria, 7-13 September 2013. Towards detecting anomalies in the content of standardized LMF-dictionaries Wafa WALI MIR@CL Laboratory, FSEGS,Sfax,Tunisia wafa.wali @fesgs.rnu.tn Bilel GARGOURI MIR@CL Laboratory, FSEGS,Sfax,Tunisia bilel.gargouri@ fesgs.rnu.tn Abdelmajid BEN HAMADOU MIR@CL Laboratory, ISIMS,Sfax,Tunisia abdelmajid.benhamadou @isimsf.rnu.tn Abstract Dictionaries are reference resources for learning and diffusing natural languages. Their contents must be enriched carefully due to their impor- tance. However, such contents might contain er- rors and inconsistencies that are hard to detect manually. Several researches have been made in recent years in order to perform this step auto- matically. However, they have dealt with the problem in a superficial way. The present paper deals with the detection of anomalies in the con- tent of LMF-standardized dictionaries that cov- ers lexical knowledge at the morphological, syn- tactic and semantic levels. Thus, we are propos- ing an approach based on a typological study of the potential anomalies that can occur in editori- al dictionaries in general. This approach takes advantage of the LMF fine structure that high- lights all kinds of relationships between entries’ knowledge and distinguishes the role of each available text such as giving definitions and ex- amples. An experiment of the proposed ap- proach was carried out on an available LMF- standardized dictionary of the Arabic language. This experiment has been related to the morpho- logical and syntactic levels. 1 Introduction Dictionaries are important linguistic resources for learning and diffusing natural languages. They can be used for several purposes such as to find the meaning, the translation, the synonym or antonym of a word. Moreover, they can help to check the spelling or to find out grammatical in- formation about a word. For ages, editorial dictionaries (for human use) have been developed in paper versions for many natural languages. With the advent of the computer science, several editorial electronic dictionaries have been constructed to be released from the constraint of their paper versions. Thus, the use of the electronic dictionaries has been expanded to meet the NLP (Natural Language Processing) needs. Then, several models have been proposed to represent the dictionary know- ledge. In addition, some projects have suggested a common representation of dictionaries such as TEI (Veronis and Ide, 1996), GENELEX (Anto- ni-Lay et al., 1994), EAGLES (Calzolari et al., 1996) and ISLE (Calzolari et al., 2003). Moreo- ver, an ISO standard has been proposed for mod- eling lexical resources and electronic dictionaries accordingly. This standard, named Lexical Mar- kup Framework (LMF: ISO 24613), provides a finely structured representation of large and common lexical knowledge (Francopoulo et al., 2008). On the other hand, a good dictionary must contain accurate knowledge to give the right an- swers for any use. Thus, it is very important to assess the quality of dictionaries’ contents, which is expensive to perform manually and requires high linguistic expertise (Fersoe and Morachina, 2004). In this context, a few works have been devoted to the evaluation of electronic dictiona- ries for many Latin and bilingual dictionaries …

Content truncated to 500 words