Using general semantic information to help the terminology structuration

We report in this paper the results of a preliminairy experience on the relevance of general semantic information in the context of a terminology structuration. As resources supplying specialised semantic information in the domain of the studied documents are missing, we use general semantic information to structure a terminology. By exploiting a general synonymy dictionary, we attempt to infer synonymy links between candidate terms extracted from French technical document. The three rules designed allow us to propose synonymy links between complex candidate terms to an expert of the domain. An interesting part of the detected links have been validated as synonymy links. On the whole, it appeared that half of the overall links is useful. We have noted also that numerous errors are due to the polysemy of general language words, coarse-grained distinction of some meanings and the elision of words widely used in the domain. Thus, we argue that general semantic information must be used with others resources such as morpho-syntactic links and is a useful aid for the terminology structuration.