Context-Sensitive Keyword Search and Ranking for XML Chavdar Botev and Jayavel Shanmugasundaram   View Paper (PDF)   Return to Poster Papers Abstract XML documents have a hierarchical structure with semantic tags, which allows users to explicitly search over a sub-set of the data. For instance, in a large heterogeneous XML digital library containing content ranging from Shakespeare's plays to scientific papers, a user who is interested in English Literature can limit the scope of a full-text search to just Shakespeare's plays. We refer to this as context-sensitive searching. Such context-sensitive searching introduces new challenges with respect to (a) efficiently evaluating keyword search queries based on a user-specified context, and (b) efficiently scoring and ranking search results. In this paper, we present novel index structures and algorithms for efficiently evaluating context-sensitive keyword search queries. We also introduce the notion of context-sensitive ranking, and develop a general ranking framework that can be tightly integrated with context-sensitive search. Finally, we present experimental results to quantify the benefits of the proposed techniques. ©2006 Association for Computing Machinery