Text & Context: towards a more semantic Information Retrieval

Abstract

With the advent of the World Wide Web, Information Retrieval has grown in significance, extending its
application beyond digital libraries to documents available on the web. Two points are essential to improve the
accuracy of the retrieval: (1) identify the concepts behind the terms that occur in the text; (2) find new concepts
semantically related to the concepts found. We show in this article which word sequences form a meaningful
term, to which disambiguation processes based on the context of the document and the neighborhood of
the term are applied. Once unambiguous, a possible spatio-temporal context of the concept is determined.
With the help of linguistic sources on the Internet, other concepts semantically related to the concept in
question are determined, and then this neighborhood network is inserted in the topic map of the document
base. The topic map thesaurus will support an adequate expansion of a query in order to accurately determine
the wanted documents..