A Method for Integrating Interfaces Based on Cluster Ensemble in Digital Library Federation

Abstract

Recently, there are more demands in a digital library federation to integrate multiple query interfaces into one for users. Since different interfaces have various descriptions for the same concept and the amount of interfaces are numerous, it is hard to provide complete and exact domain knowledge. Hence, the methods of clustering are usually adopted to generate an integrated interface. However, over one same properties set, the results for clustering may be diverse according to the differences of clustering algorithms or parameters setting for the same algorithm. Nevertheless, we could obtain one more complete and exact integrated interface with the aid of cluster ensemble by merging multiple clustering results. In this paper, based on the principle of cluster ensemble, we propose a single clustering algorithm with uncertainty regarding that one property may belong to more than one possible cluster division during integration. We also propose a fusing cluster algorithm to obtain cluster ensemble that satisfying interface integration and it shows favorable performances than the existing methods.