Research On Approximate Matching Model For XML-based Metadata

As the base of description and retrieval of resource objects, XML and metadata technologies have been widely researched and applied in all kinds of fields. Though many experts have especially focused on the study of retrieval technology on XML and metadata, but still no efficient retrieval methods could ensure the accuracy and recall rate. This thesis aims at the efficiency problem of accuracy and recall rate in XML retrieval. Based on tree matching, the thesis makes a further study on the XML retrieval technology about metadata.This thesis introduces the basic theory and method about XML information retrieval, and studies the development of XML retrieval methods, especially about the tree matching theory and relative algorithms.In order to improve the recall rate in XML retrieval, this thesis takes apart the unordered label tree matching into tree structure matching and tree label semantic matching. This thesis changes the classical tree matching algorithms into approximate matching by combining the tree structure matching with semantic matching. According to the tree characteristics of metadata description, it puts forward a metadata retrieval method based on tree approximate matching and gave out a new retrieval method based on three-level tree approximate matching model. This retrieval method can adjust the accuracy and recall rate of searching metadata by different users.Because of using the conception of edition distance and mapping cost and the method of dynamic programming, the thesis analyzes the counting question of approximate matching cost in detail. Due to the calculation for the edition distance of unordered label tree belongs to NP-Hard problem, according to the structured features of XML metadata searching in the above match models, this thesis adds into some proper limiting conditions, so that the calculation of the approximate matching cost can be solved in multinomial time.In the end, using the above approximate matching model and the calculation of matching cost, this thesis puts forward the retrieval process based on the XML description metadata, and designs the architecture of query system for XML resource objects. The experimental results prove that the approximate matching model is feasible and efficient in the practical application of metadata retrieval.