A structure-based filtering method for XML management systems

Abstract

To answer queries, many XML management systems perform structural joins, i.e., they determine all occurences of parent/child or ancestor/descendant relationships between node sets. These joins are often one of the most time-consuming phases in query evaluation, so it is desirable to reduce the size of the node sets before performing the joins. This problem has earlier been approached by using signatures built on the content of the nodes, but in this paper, we propose a novel method in which the nodes are filtered based on the structural properties of their subtrees. To achieve this, we use a schema graph which summarizes the structures of XML documents more accurately than conventional summarization methods. © Springer-Verlag Berlin Heidelberg 2004.