A Common Access Structure for Standard Attributes and Document Representations in Vector Space

In next generation information systems there will be a coalescence of (object-oriented) database management systems and information retrieval systems. Especially the integration of content-based retrieval techniques into object-oriented database management systems is an interesting requirement in this respect. Whereas the aspects of this coalescence dealing with the query language and the data model have been addressed in some recent papers, approaches dealing with the integration at the physical level are missing. In the present paper we propose a common access structure which can support a content-based similarity search with additional conditions on standard attributes in one homogeneous step. To this end, we use a k-dtree based multi-attribute access structure which considers standard attributes in the rst dimensions and the components of a document description vector in the higher dimensions. We describe the algorithms for this access structure and present rst performance results.