Heterogeneous Web Data Search Using Relevance-based On The Fly Data Integration
Published: 2012 April
Buchtitel: World Wide Web Conference (WWW2012)
Erscheinungsort: Lyon, France
Searching over heterogeneous structured data on the Web is challenging due to vocabulary and structure mismatches among different data sources. In this paper, we study two main directions. The first one relies on data integration to mediate these mismatches through upfront computation of mappings, based on which queries are rewritten to fit the vocabulary and structure of individual sources. The other extreme is keyword search, which does not require any up-front investment, but ignores structure information that can be exploited for more effective search. Then, we present a hybrid approach, which assumes only one single structured query that adheres to the vocabulary of just one of the sources. However, this so-called seed query is not rewritten to obtain structured queries for individual sources, but processed as a keyword query. For more effective keyword search that also takes structure information into account, we construct an entity relevance model (ERM), which captures both the content and structure of the seed query results. On the fly, this ERM model is then aligned with keyword search results retrieved from other sources to bridge vocabulary mismatches, and finally used to rank these results. Through experiments using large-scale real world datasets, we study these three different strategies. The outcomes suggest that upfront investment in data integration leads to higher search effectiveness compared to keyword search, and that the hybrid strategy clearly provide best results.