Query Relaxation for XML Databases

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

Due to the expressiveness of Extensible Markup Language (XML) as compared to HTML, XML has become a standard format in data presentation for the web applications. However, in content based retrieval, if the database content does not match with the query conditions exactly, or if the query conditions are not well presented, the system returns null answer. The user must reformulate and resubmit the query for processing. Therefore, when the query conditions are not known exactly (e.g. matching an image with a set of features or a document with a set of key words), the user often specifies them in a general way to avoid a null answer. The drawback of such an approach is that the system may return too many and sometimes irrelevant answers. To remedy this problem, a systematic and scalable knowledge based (derived from the XML data) relaxation methodology will be developed to relax the query conditions (both in value and / or structure). Further, the user can provide such relaxation control as the order of attributes to be relaxed, relaxation level, the unrelaxable attribute list, number of answers, etc. As a result, such technique provides a focused and controlled relaxation, and thus yields more accurate and relevant answers. In this proposal, the existing XML query language will be extended to support query relaxation and relaxation control constructs, and techniques will be developed to cluster similar data trees from XML data into a hierarchy to serve as an index to guide relaxation.