Arabic/English word translation disambiguation using parallel corpora and matching schemes

Arabic/English Word Translation Disambiguation using Parallel Corpora and Matching Schemes Farag Ahmed and Andreas Nürnberger Data and Knowledge Engineering Group Faculty of Computer Science Otto-von-Guericke-University of Magdeburg Building 29, Universitätsplatz 2 39106 Magdeburg {farag.ahmed, andreas.nuernberger}@ovgu.de Abstract. The limited coverage of available Arabic language lexicons causes a serious challenge in Arabic cross language information retrieval. Translation in cross language information retrieval consists of assigning one of the semantic representation terms in the target language to the intended query. Despite the problem of the completeness of the dictionary, we also face the problem of which one of the translations proposed by the dictionary for each query term should be included in the query translations. In this paper, we describe the implementation and evaluation of an Arabic/English word translation disambiguation approach that is based on exploiting a large bilingual corpus and statistical co-occurrence to find the correct sense for the query translations terms. The correct word translations of the given query term are determined based on their cohesion with words in the training corpus and a special similarity score measure. The specific properties of the Arabic language that frequently hinder the correct match are taken into account. Keywords: Word Translation Disambiguation (WTD), Arabic language, Parallel Corpus, Naïve Bayesian Classifiers 1 Introduction The meaning of a word may vary significantly according to the context in which it occurs. As a result, it is possible that some words can have multiple meanings. This problem is even more complicated when those words are translated from one language into others. Therefore there is a need to disambiguate the ambiguous words that occur during the translations. The word translation disambiguation, in general, is the process of determining the right sense of an ambiguous word given the context in which the ambiguous word occurs. We can define this word sense disambiguation (WSD) problem, as the association of an occurrence of an ambiguous word with one of its proper sense. Diacritization in Arabic (sometimes called vocalization or voweling) can be defined as a symbol over and underscored letters, which are used to indicate the proper pronunciations as well as for disambiguation purposes. The absence of diacritization in Arabic texts poses a real challenge for Arabic natural language 12th EAMT conference, 22-23 September 2008, Hamburg, Germany 6 processing as well as for translation, leading to high ambiguity the absence of the diacritization in most of the Arabic printed media or on the Internet web sites lead to high ambiguity. Thus, the probability that a single word can have multiple meanings is even a lot higher. For example, the Arabic word “ﺪﻌﻳ“ can have these translations in English (Promise, Prepare, count, return, bring back) or the Arabic word “ﻢﻠﻋ“ can have these possible translations (flag, science, he knew, it was known, he taught, he was taught). The task of disambiguation therefore involves two processes: Firstly, identifying all senses for every word relevant, secondly assign each time this word occurs the appropriate sense to it. For the first step, this can be done …

Content truncated to 500 words