Dynamically Bounding Browsable Retrieval Spaces: an Application to Galois Lattices

In this paper we present a method that allows the user to change an ordered retrieval space while browsing through it. Based on the feedback the user gets from the space structure s/he may dynamically assert or retract constraints which the sought classes of documents have to comply with while the space is bounded accordingly using a representation called version space. We apply the method to a Galois lattice space - a classification stucture that has recently gained some attention for its theoretical and practical advantages for browsing retrieval - and we analyse the improved retrieval capabilities of the integrated framework. The bound method has been implemented as a component of a larger retrieval interface for browsing and querying the Galois lattice of a bibliographic database, possibly enriched with a thesaurus of terms.