Interactive search over Web scale RDF data using predicates as constraints

Abstract

RDF (Resource Description Framework) data are more and more prevalent in the applications of semantic web and web data publication. The search over Web scale RDF data is essential for users to retrieve desired information from the huge RDF datasets, which typically applied as knowledge bases supporting many advanced information seeking tasks. In this paper, we propose some techniques that allow users to interactively search over the Web scale RDF data by using keywords as well as their predicates as additional constraints. We observe that the straightforward way of keyword search over the Web scale RDF data often generates a huge number of matching sub-structures (i.e., graphs containing the query keywords) due to the ambiguity of query intention generated from a small number of query keywords, although most of them are false interpretations of the query intention. To effectively interpret the semantics of queries, we define a novel keyword query called structure-aware keyword query that utilizes the predicates of RDF triples to assist users in clarifying their query intention. The challenge of such queries is to effectively and efficiently find a proper set of predicate-keyword pairs for query interpretation, to reduce the manual cost of user feedbacks as much as possible. To verify the novel query mechanism, we implement a system, and test it over the DBPedia 3.7 dataset. Results show that, for most queries, users can often efficiently get desired results by providing a small number of simple feedbacks on the constraints of predicates automatically generated.