Flexible Similarity Assessment for XML Documents Based on XQL and Java Reflection

In this paper we present a flexible similarity assessment framework for XML documents and describe its usage on the exam- ple of XML-based fault diagnosis for remote fieldbus automation sys- tems. Our approach combines the expressive power of both XQL and Java to overcome the limitations of standard XML query languages and to provide a convenient platform for rapid definition of distance func- tion for any kind of XML structured data. The presented framework allows the extraction of fragments from XML documents, the wrapping of these fragments in any desired Java objects, the dynamic invocation of Java methods with the created objects as parameters and the com- putation of a scalar overall result from the partial results which can be interpreted as a distance value. The actual mapping of XML fragments to Java instances assessing their similarity is specified by an XML to Java Mapping Language (XJML) XML document in a formal and flexi- ble way. The framework consists of the XML to Java mapping concept, the XJML DTD, and various Java packages to evaluate an XJML-based nearest neighbor search on local documents or on XQL query result sets of (remote) XML database systems.