Cooperative XML (CoXML) query answering

The growing use of XML in scientific data repositories, digital libraries and Web applications has increased the need for effective XML search methods. Due to the heterogeneous nature of XML data, it is often difficult and unrealistic for users to completely grasp the structure properties of the data before asking queries. As a result, it may be often the case that there are no exactly matched answers for a query. In this case, it is more cooperative to derive approximate answers than to return users an empty answer set. In this dissertation, we propose a cooperative XML (CoXML) query answering system that relaxes query conditions to less restricted forms, i.e., query relaxation, and derives approximate answers.; The key features of the system include: (1) a query language that extends the standard W3C XQuery with relaxation constructs and controls. Relaxation constructs enable users to specify approximate conditions and relaxation controls allow users to control the approximate matching process; (2) a multi-level hierarchical relaxation index structure that provides systematic approximate matching guidance and controls at different granularity; and (3) similarity metrics that evaluate the relevancies of approximate answers based on both their structure and content aspects.; We evaluate the system's performance by using the INEX test collection with its relevant assessments as the "gold standard." The results reveal that the proposed relaxation constructs and controls allow users to effectively express their approximate matching specifications, which in turn enable the system to provide answers with more accuracy. The results also demonstrate the effectiveness of the proposed content and structure similarity metrics in result ranking. Furthermore, compared to other systems participated in the INEX, our query relaxation methodology not only enables the system to systematically retrieve approximate answers but also with greater relevancy.