A Possibilistic Approach for Automatic Word Sense Disambiguation

1 A possibilistic approach for automatic word sense disambiguation Oussama Ben Khiroun1, Bilel Elayeb1,2, Ibrahim Bounhas3, Fabrice Evrard2, and Narjès Bellamine Ben Saoud1 1RIADI Research Laboratory, ENSI Manouba University, 2010 Tunisia. oussama.ben.khiroun@gmail.com, Bilel.Elayeb@riadi.rnu.tn, Narjes.Bellamine@ensi.rnu.tn 2IRIT-ENSEEIHT, 02 Rue Camichel, 31071 Toulouse Cedex 7, France. Fabrice.Evrard@enseeiht.fr 3Department of Computer Science, Faculty of Sciences of Tunis, 1060 Tunis, Tunisia Bounhas.Ibrahim@yahoo.fr Abstract This paper presents and experiments a new approach for automatic word sense disambiguation (WSD) applied for French texts. First, we are inspired from possibility theory by taking advantage of a double relevance measure (possibility and necessity) between words and their contexts. Second, we propose, analyze and compare two different training methods: judgment and dictionary based training. Third, we summarize and discuss the overall performance of the various performed tests in a global analysis way. In order to assess and compare our approach with similar WSD systems we performed experiments on the standard ROMANSEVAL test collection. Keywords: Word Sense Disambiguation, Semantic Dictionary of Contexts, Possibility Theory. 1. Introduction Word Sense Disambiguation (WSD) is the ability to identify the meaning of a word in its context in a computational manner. A lexical semantic disambiguation allows to select in a predefined list the significance of a word given its context. In fact, the task of semantic disambiguation requires enormous resources such as labeled corpora, dictionaries, semantic networks or ontologies. This task is important in many fields such as optical character recognition, lexicography, speech recognition, natural language comprehension, accent restoration, content analysis, content categorization, information retrieval and computer aided translation [13] [14]. The problem of WSD has been considered as a difficult task in the field of Natural Language Processing. In fact, a reader is frequently faced to problems of ambiguity in information retrieval or automatic translation tasks. Indeed, the main idea on which were based many researches in this field is to find relations between an occurrence of a word and its context Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012) 261 2 which will help identify the most probable sense of this occurrence [1][2]. We discuss in this paper the contribution of a new approach for WSD. We presuppose that combining knowledge extracted from corpora and traditional dictionaries will improve disambiguation rates. We also show that this approach may perform satisfactory results even without using manually labeled corpora for training. We also propose to apply possibility theory as an efficient framework to solve the WSD problem seen as a case of imprecision. Indeed, WSD approaches need training and matching models which compute the similarities (or the relevance) between senses and contexts. Existing models for WSD are based on poor, uncertain and imprecise data. Whereas, possibility theory is naturally designed to this kind of applications; because it makes it possible to express ignorance and to take account of the imprecision and uncertainty at the same time. For example a recent work of Ayed et al. (2012) [23][24] which have proposed possibilistic approach for the morphological disambiguation of arabic texts showed …

Content truncated to 500 words