Suggestion of promising result types for XML keyword search

Although keyword query enables inexperienced users to easily search XML database with no specific knowledge of complex structured query languages or XML data schemas, the ambiguity of keyword query may result in generating a great number of results that may be classified into different types. For users, each result type implies a possible search intention. To improve the performance of keyword query, it is desirable to efficiently work out the most relevant result type from the data to be retrieved.

Several recent research works have focused on this interesting problem by using data schema information or pure IR-style statical information. However, this problem is still open due to some requirements. (1) The data to be retrieved may not contain schema information; (2) Relevant result types should be efficiently computed before keyword query evaluation; (3) The correlation between a result type and a keyword query should be measured by analyzing the distribution of relevant values and structures within the data. As we know, none of existing work satisfies the above three requirements together. To address the problem, we propose an estimation-based approach to compute the promising result types for a keyword query, which can help a user quickly narrow down to her specific information need. To speed up the computation, we designed new algorithms based on the indexes to be built. Finally, we present a set of experimental results that evaluate the proposed algorithms and show the potential of this work.