MARS: A Matching and Ranking System for XML Content and Structure Retrieval

Structural queries specify complex predicates on the content and the structure of the elements of tree-structured XML documents. Recent works have typically applied top-down decomposition of the twig patterns into (i) parent-child or ancestordescendant relationships, or (ii) path expression queries, and then followed by a join operation to reconstruct matched twig patterns. This demonstration system is the implementation prototype of an efficient heuristic-based bottom-up approach named MARS (Matching And Ranking System for xml structure queries), for matching and ranking of structural query patterns for XML query processing. An efficient nearest common ancestor labeling scheme is applied to utilize fast bottomup construction of the subtree matches from the potential keywords. MARS considers both the content and structure of queries and incorporates a variation of IR-based relevance ranking to report the top-k ranked results. The graphical user interface of MARS provides an interactive visualization of the twig query discovery. 1 Motivation of MARS Structured twig queries specify complex patterns of selection predicates on element labels (keyword search) and their corresponding inherent structural relationships. XML query languages [1, 2, 3, 4] and the underlying relational [5, 7] or native [6, 8] databases must provide efficient and effective support for querying both the content and the inherent structure of XML documents. However, the support for structural queries is usually facilitated by augmenting a layer of structural search on top of the traditional path expression query language performing the following general steps: (i) Decompose the twig query structure into its corresponding path expression queries. For instance, (/dblp/ /journal[author=’Serge’ AND title=’XML’]) gets decomposed into (/dblp//journal [author=’Serge’]) and ( /dblp//journal[title=’XML’]) path queries, (ii) Perform Twig Matching & Ranking 1