XSnippets: Exploring semi-structured data via snippets

Abstract

Users are usually not familiar with the content and structure of the data when they explore the data source. However, to improve the exploration usability, they need some primary hints about the data source. These hints should represent the overall picture of the data source and include the trending issues that can be extracted from the query log. In this paper, we propose a two-phase interactive exploratory search framework for the clueless users that exploits the snippets for conducting the search on the XML data. In the first phase, we present the primary snippets that are generated from the keywords of the query log to start the exploration. To retrieve the primary snippets, we develop an A* search-based technique on the keyword space of the query log. To improve the performance of computations, we store the primary snippet computations in an index data structure to reuse it for the next steps. In the second phase, we exploit the co-occurring content of the snippets to generate more specific snippets with the user interaction. To expedite the performance, we design two pruning techniques called inter-snippet and intra-snippet pruning to stop unnecessary computations. Finally, we discuss a termination condition that checks the cardinality of the snippets to stop the interactive phase and present the final Top-l snippets to the user. Our experiments on real datasets verify the effectiveness and efficiency of the proposed framework.