Retrieving Top Relaxed Tightest Fragments for XML Keyword Search

In this paper, we revisit the problem of retrieving meaningful result for XML keyword search. We begin by introducing the concept of Relaxed Tightest Fragment (RTF) as the basic result type, which preserves all the related keyword nodes implied in corresponding fragments in the XML data. This is different from the prior popular proposals like LCA (Lowest Common Ancestor), SLCA (Smallest LCA) variants. We then propose a ranking method, XKSMetaRank, to compute the meaningfulness of the RTFs, which takes into account the information contained both in the meta data and the XML data. Finally, we propose a flexible algorithm to cope with this problem, named as LISA/RTF (Layered Intersection Scan Algorithm). It takes a bottom-up layer-by-layer way to collect the RTFs and computes its meaningfulness. We implement our algorithm, and compare it with the method in [1]. Experiments on real and synthetic data verify the efficiency and effectiveness of our method.