Improving the Effectiveness of XML Retrieval with User Navigation Models

Structured documents (predominantly encoded in XML) utilize markup dialects for several purposes, such as conveying logical structure, or providing rendering instructions. XML structure can also help users to navigate within documents to satisfy their information needs. However, including the user's structural preferences in the ranking of retrieved elements remains a key challenge in XML retrieval. In this paper, we propose an approach for including structural preferences in the ranking of XML elements by improving the structural relevance (SR) of results. SR is an evaluation measure which relies on graphical navigation models to capture the structural preferences of users. We propose several algorithms to post-process search engine output to improve the SR of the output. Experimental results (using data, assessments, and search engines from INEX 2007 and 2008) demonstrate the effect of different combinations of post-processing algorithms and navigation models on the effectiveness of systems.