Cross-language information retrieval using meta-language index construction and structural queries

Abstract

Structural Query Language allows expert users to richly represent its information needs but unfortunately, the complexity of SQLs make them impractical in the Web search engines. Automatically detecting the concepts in an unstructured user's information need and generating a richly structured, multilingual equivalent query is an ideal solution. We utilize Wikipedia as a great concept repository and also some state of the art algorithms for extracting Wikipedia's concepts from the user's information need. This process is called "Query Wikification". Our experiments on the TEL corpus at CLEF2009 achieves +23% and +17% improvement in Mean Average Precision and Recall against the baseline. Our approach is unique in that, it does improve both precision and recall; two pans that often improving one, hurt the another. © 2010 Springer-Verlag Berlin Heidelberg.