Exploiting semantics for XML keyword search

XML keyword search has attracted a lot of interests with typical search based on lowest common ancestor (LCA). However, in this paper, we show several problems of the LCA-based approaches, including meaningless answers, incomplete answers, duplicated answers, missing answers, and schema-dependent answers. To handle these problems, we exploit the semantics of object, object identifier, relationship, and attribute (referred to as the ORA-semantics). Based on the ORA-semantics, we introduce new ways of labeling and matching. More importantly, we propose a new semantics, called CR (Common Relative) for XML keyword search, which can return answers independent from schema designs. To find answers based on the CR semantics, we discover properties of common relative and propose an efficient algorithms. Experimental results show the seriousness of the problems of the LCA-based approaches. They also show that the CR semantics possesses the properties of completeness, soundness and independence while the response time of our approach is faster than the LCA-based approaches thanks to our techniques.