Approximate Joins for XML Using g-String

Abstract

When integrating XML documents from autonomous databases, exact joins often fail for the data items representing the same real world object may not be exactly the same. Thus the join must be approximate. Tree-edit-distance-based join methods have high join quality but low efficiency. Comparatively, other methods with higher efficiency cannot perform the join as effectively as tree edit distance does.

To keep the balance between efficiency and effectiveness, in this paper, we propose a novel method to approximately join XML documents. In our method, trees are transformed to g-strings with each entry a tiny subtree. Then the distance between two trees is evaluated as the g-string distance between their corresponding g-strings. To make the g-string based join method scale to large XML databases, we propose the g-bag distance as the lower bound of the g-string distance. With g-bag distance, only a very small part of g-string distance need to be computed directly. Thus the whole join process can be done very efficiently. We theoretically analyze the properties of the g-string distance. Experiments with synthetic and various real world data confirm the effectiveness and efficiency of our method and suggest that our technique is both scalable and useful.

Supported by the National Science Foundation of China (No 60703012, 60773063), the NSFC-RGC of China(No. 60831160525), National Grant of Fundamental Research 973 Program of China (No.2006CB303000), National Grant of High Technology 863 Program of China (No. 2009AA01Z149), Key Program of the National Natural Science Foundation of China (No. 60933001), National Postdoctor Foundtaion of China (No. 20090450126), Development Program for Outstanding Young Teachers in Harbin Institute of Technology (no. HITQNJS.2009.052).