Efficient similarity search in digital libraries

Abstract

Digital libraries are a core information technology. When the stored data is complex, e.g. high-resolution images or molecular protein structures, simple query types such as exact match query are hardly applicable. In such environments similarity queries, particularly range queries and k-nearest neighbor queries, are important query types. Numerous approaches have been proposed for the processing of similarity queries which mainly concentrate on highly dynamic data sets where insertion, update, and deletion operations occur. However, only little effort has been devoted to the case of rather static data sets-frequently, occurring in digital libraries. In this paper we introduce a novel technique for efficient similarity search on top of static or rarely changing data sets. In particularly we propose a special sorting order on the data objects which can be effectively exploited to substantially reduce the total query time of similarity queries. An extensive experimental evaluation with real-world data sets emphasizes the practical impact of our technique.