An Effective Object-Level XML Keyword Search

Keyword search is widely recognized as a convenient way to retrieve information from XML data. In order to precisely meet users' search concerns, the authors study how to effectively return the targets that users intend to search for. They model XML document as a set of interconnected object-trees, where each object contains a subtree to represent a concept in the real world. Based on this model, the authors propose object-level matching semantics called Interested Single Object (ISO) and Interested Related Object (IRO) to capture single object and multiple objects as user's search targets respectively, and design a novel relevance oriented ranking framework for the matching results.