The work presented in this paper addresses the problem of interpretation and semantic classification of documents. One of the issues faced by natural languages is related to the presence, in glossaries, of words with similar morphologies and different meanings. Our approach is based on the use of domain ontologies for nouns disambiguation. We begin our process with a global disambiguation, by linking the considered document to a semantic domain (represented by an ontology) which we select among several candidate ones. We define a candidate domain as any domain in which at least one significant word of the text can be considered and makes sense. We then perform a local disambiguation by using the selected ontology and finally build a semantic representation of the content of the document as a conceptual graph.
展开▼
