Building domain specific lexical hierarchies from corpora
Proceedings of the Third International Conference on Language Resources and Evaluation (LREC 2002)
Abstract
In this article, we present a new algorithm for building domain specific lexical hierarchies from texts. The basic elements of such a hierarchy are the normalized terms - mono and multi-word terms - extracted from a large corpus by a terminological extractor. The algorithm relies on collocations for representing the meaning of these terms, finding hierarchical relations between them and finally, organizing them into a hierarchy. Moreover, it takes into account the polysemy of terms while it builds the hierarchy. We also present the results of its application on a part of the corpus designed for the ARC A3 of the Francil network and we go through its possible applications.