Efficient Parallel Algorithms for XML Filtering with Structural and Value Constraints

Abstract

Information seeking applications employ information filtering techniques as a main component of their functioning. The purpose of the present article is to explore techniques to efficiently implement scalable and efficient information filtering techniques, based on the XML representation, when both structural and value constraints are imposed. In the majority of the provided implementations the use of the XML representation appears in single processor systems, and the involved user profiles are represented using the XPath query language and efficient heuristic techniques for constraining the complexity of the filtering mechanism are employed. Here, we propose a parallel filtering algorithm based on the well known YFilter algorithm, which dynamically applies a work-load balancing approach to each thread to achieve the best parallelization. In addition, the proposed filtering algorithm adds support for value-based predicates by embedding three different algorithms for handling value constraints during XML filtering, based on the popularity and the semantic interpretation of the predicate values. Experimental results depict that the proposed system outperforms the previous parallel approaches to the XML filtering problem.