Friendly information retrieval through adaptive restructuring of information space

Abstract

Although relevance feedback techniques are relatively common in the field of information retrieval (IR), feedback usually supports a process of query refinement. Using feedback to restructure the information space itself has yet to be attempted. Restructuring not only supports useful applications such as clustering, but is also indispensable for IR given that the modeling function employs inter-term correlation. This paper presents a new approach to relevance feedback involving information space manipulation, and examines its effectiveness through a number of experiments.