Prefiltering techniques for efficient XML document processing

Document Object Model (DOM) and Simple API for XML (SAX) are the two major programming models for XML document processing. Each, however, has its own efficiency limitation. DOM assumes an in-core representation of XML documents which can be problematic for large documents. SAX needs to scan over the document in a linear manner in order to locate the interesting fragments. Previously, we have used tree-to-table mapping and indexing techniques to help answer structural queries to large, or large collections of, XML documents. In this paper, we generalize the previous techniques into a prefiltering framework where repeated access to large XML documents can be efficiently carried out within the existing DOM and SAX models. The prefiltering framework essentially uses a tiny search engine to locate useful fragments in the target XML documents by approximately executing the user's queries. Those fragments are gathered into a candidate-set XML document, and is returned to the user's DOM- or SAX-based applications for further processing. This results in a practical and efficient model of XML processing, especially when the XML documents are large and infrequently updated, but are frequently being queried.