XEdge: An Efficient Method for Returning Meaningful Clustered Results for XML Keyword Search

Abstract

In this paper, we investigate the problem of returning meaningful clustered results for XML keyword search. We begin by presenting a multi-granularity computing methodology, in order to make full use of the structural information of XML trees to extract features. In this method, we first propose the concept of Cluster Compactness Granularity (CCG) to partition the search results into different clusters, which enable users to precisely and quickly seek their desired answers, according to the connection compactness between LCA nodes. We then propose the concept of Subtree Compactness Granularity (SCG) to rank individual results within clusters and measure the query result relevance. Furthermore, we define a novel semantics of Compact LCA (CLCA), which not only improves the accuracy by eliminating redundant LCAs that do not contribute to meaningful answers, but also overcomes the shielding effects of SLCA-based methods. Using the proposed CCG and SCG features and the CLCA semantics, we finally implement an efficient algorithm called XEdge for generating meaningful clustered results. Comparing with the existing methods such as XSeek and XKLUSTER, the experimental results demonstrate the effectiveness of the proposed multi-granularity clustering methodology and validity of the complemented ranking strategy, as well as the meaningfulness of CLCA semantics.

This work was partially supported by NSFC (No. 61272374, 61300190), Program for NCET in University of China (No. NCET-11-0056), Specialized RFDP of Higher Education (No.20120041110046), Key Project of Chinese Ministry of Education(No. 313011) and the Fundamental Research Funds for the Central Universities (No. DUT13JR04).