Special Topic Data Base Development

Abstract

Two basic approaches to the development of special topic document corpora are considered. Both techniques result in a partitioning of a general heterogeneous parent corpus into distinct subsets of related documents. The first method is based on the partitioning of a document corpus with respect to the topical content explicitly defined by the documents contained in the corpus. The technique relies on a logicosyntactic analysis of the document text in order to extract topic-denoting phrases, and a weighting function based on the complexity of the logical relational environment of the extracted phrases. The second method is based on a profile-directed partitioning of the document corpus induced by an externally defined thesaurus of phrases. The topic coverage of the profile depends only on the specific requirements of the user community for whom it was defined. Any one of a number of weighting functions can be applied to the phrases and usually depends on the corpus itself. This technique is useful where text analysis is either impractical or not possible.