Distributions

Amsterdam: IOS Pressidentify translation equivalents at the word- or phrase level. Such techniques can be useful especially for medical and techni-cal domains which are in constant evolution, producing new terms as knowledge advances. Many related works on using sta-tistical models for mapping bilingual terms [1, 2] are based on parallel texts or ‘bitexts ’ —pairs of texts that are translation of each other. Most of these methods are based on the following as-sumption: words that are translations of each other are more like-ly to appear in corresponding parallel text regions than other pairs of words. By using various correlation metrics, these ap-proaches derive co-occurrence patterns of words across languag-search on alignment of non-parallel texts at the word level and to research on domain-specific bilingual lexicon acquisition. Previ-ous works in this area are based on the assumption that words which have the same meaning in different languages should have similar context distributions. Rapp [5] proposes an approach very similar to the method pre-sented here. He supposes that in any language there is a correla-tion between the cooccurrences of words which are translations of each other. The main difference between our approach and his model is that he used lemmatized corpora and limited the num