G. Ramirez and A.P de Vries

Combining Indexing Schemes to Accelerate Querying XML on Content and Structure


This paper presents the advantages of combining multiple document representation schemes for query processing of XML queries on content and structure. We show how extending the Text Region approach [2] with the main features of the Binary Relation approach developed in [8] leads to a considerable speed-up in the processing of the XPath location steps. We detail how, by using the combined scheme, we reduce the number of structural joins used to process the XPath steps, while simultaneously limiting the amount of memory usage. We discuss optimisation strategies enabled by the new `combined representation scheme'. Experiments comparing the efficiency of alternative query processing strategies on a subset of the queries used at INEX 2003 (the Initiative for the Evaluation of XML Retrieval [4]) demonstrate a favourable performance for the combined indexing scheme.