SemanticTwig: A Semantic Approach to Optimize XML Query Processing

Abstract

Twig pattern matching (TPM) is the core operation of XML query processing. Existing approaches rely on either efficient data structures or novel labeling/indexing schemes to reduce the intermediate result size, but none of them takes into account the rich semantic information resided in XML document and the query issued. Moreover, in order to fulfill the semantics of the XPath/XQuery query, most of them require costly post processing to eliminate redundant matches and group matching results. In this paper, we propose an innovative semantics-aware query optimization approach to overcome these limitations. In particular, we exploit the functional dependency derived from the given semantic information to stop query processing early; we distinguish the output and predicate nodes of a query, then propose a query breakup technique and build a query plan, such that for each distinct query output, we avoid finding the redundant matches having the same results as the first match in most cases. Both I/O and structural join cost are saved, and much less intermediate results are produced. Experiments show the effectiveness of our optimization.