Acquiring Lexical Classes in Biomedical Lexicons: a Morphosemantics-based Multilingual Approach.
Résumé
The aim of this paper is to present a multilingual method for the structuration of biomedical lexicons into semantic networks. The method is based on the use of a standard classification model in biomedecine: that of UMLS (Unified Medical Language System). The approach results in the acquisition of semantic information on morphologically complex lexical units, and, more especially, in that of 3 sorts of lexical relations between so-called neoclassical compound nouns and adjectives: synonymy, hyponymy, co-hyponymy. The interest of this approach is related to the continuous increasing of biomedical vocabulary: as noticed in [1], it is impossible for medical dictionaries to collect all neologisms. Now, more than 60% of these unknown words are morphologically constructed words - affixed or compound -, so that their meaning is partially calculable by means of that of their morphological base or components. Moreover, very often and whatever the language (see [2]), these bases or components are so-called combining forms (henceforth CFs), i.e. non autonomous latin or greek elements found either in derived words: epaticoit, stomacalfr or as compound elements: phyto/phagousen, lipó/lisises