Implementation of a stream-oriented retrieval engine for complex similarity queries on top of an ORDBMS

Abstract

When dealing with structured multimedia documents the typical query is no longer an exact match query, but a best match or similarity query yielding a ranking for the required objects. To process such queries different components are needed -namely, rankers delivering a sorting of objects of a given type with respect to a single similarity criterion, combiners merging multiple rankings over the same set of objects and transferers transferring a ranking for objects of a given type to related objects. In the literature various approaches for these single components have been presented. However, the integration of the components into a comprehensive approach for complex similarity queries has hardly been addressed. In this paper we propose IRstream as a retrieval engine for the stream-oriented processing of complex similarity queries. This retrieval engine is intended to complement traditional query processing techniques for queries dominated by similarity conditions. It utilizes rankers, combiners and transferers and it is implemented on top of an ORDBMS. We describe the concept and the architecture of the system and state some experimental results. © Springer-Verlag Berlin Heidelberg 2003.