Filtering contents by efficient evaluation of XPath expressions

The need for selection of contents based on personal data is increasing. Filtering systems using XML documents as users' profiles and XPath as contents' conditions are developed. They need efficient algorithms for evaluating a large number of XPath. However existing algorithms have limitations in that (1) they cannot handle the full XPath language because their execution models are event-based, (2) they require time proportional to the number of XPath expressions, and (3) they do not allow us to add or remove XPath expressions incrementally. We present algorithms that solve all of these problems. The main idea of the algorithms is (1) to construct a data structure that represents multiple XPath expressions in order to evaluate shared subexpressions only once and (2) to repeatedly apply a conventional XPath processor to the data structure. We show that our algorithm is efficient by evaluating the performance of a prototype implementation.