Predication-based semantic indexing: permutations as a means to encode predications in semantic space.

A novel vector space model is presented that encodes semantic predications derived from MEDLINE by the SemRep system into a compact spatial representation and presents new possibilities for knowledge discovery and information retrieval.

Abstract

Corpus-derived distributional models of semantic distance between terms have proved useful in a number of applications. For both theoretical and practical reasons, it is desirable to extend these models to encode discrete concepts and the ways in which they are related to one another. In this paper, we present a novel vector space model that encodes semantic predications derived from MEDLINE by the SemRep system into a compact spatial representation. The associations captured by this method are of a different and complementary nature to those derived by traditional vector space models, and the encoding of predication types presents new possibilities for knowledge discovery and information retrieval.