FIX: feature-based indexing technique for XML documents
Ning Zhang, M. TAMER ÖZSU, Ihab F. Ilyas, Ashraf Aboulnaga · 2006
In this paper, we study the problem of indexing an XML database. Existing XML indexing techniques focus on clustering methods based on the combinatorial structural properties of an XML document. These techniques cluster tree nodes into an index tree or graph based on their similarities in ancestor-descendant or sibling relationships. Index look-up then amounts to pattern matching on the clustered tree or graph. In this paper, we propose a feature-based indexing technique, called FIX, based on the spectral graph theory. The basic idea is that for each twig pattern in a collection of XML documents, we calculate a vector of features based on its structural properties. These features are used as a key for the patterns and stored in a B-tree or a multidimensional index tree. Given an XPath query, its feature vector is first calculated and looked up in the index. Then a further refinement phase is performed to fetch the final results. We experimentally study the indexing technique over two scenarios: a large collection of relatively smaller documents, and a single large document. Our experiments show that FIX provides great pruning power and could gain an order of magnitude performance improvement for many XPath queries over existing evaluation techniques.