Accessing Information Based on a Combination of Document Structure and Content: Exploiting XML tags in indexing and searching to enhance content retrieval of online document-centric XML encoded texts
This study explores the challenges of using traditional information retrieval methods to retrieve document-centric XML encoded text. It demonstrates how coupling structure and content in query and index formulation improves retrieval performance. Native XML database (NXD) and search engine technologies were evaluated in a baseline experiment, and in a second test after alterations were made to their respective indexes. Documents were retrieved for simple and complex forms of 30 XPath and keyword queries from a corpus of 95 XML/TEI encoded texts. Overall results indicated that query augmentation using document structure improves retrieval performance. Complex queries submitted to the NXD produced the most satisfying results, with an average precision of 93.3% and an average recall of 86.3%. Performance improvements were also achieved using complex, structured queries and indexes in the search engine. Study findings suggest that effective XML retrieval models might result from a combination of unstructured and structured retrieval techniques.
Advisor:Stephanie W. Hass
School:University of North Carolina at Chapel Hill
School Location:USA - North Carolina
Source Type:Master's Thesis
Keywords:information retrieval xml search semistructured data indexing full text searching
Date of Publication:04/12/2004