Context Awareness On The Rdf Graphs Of Keywords Retrieval

At present, increasing web resources are being described by using RDF. Keyword search over RDF, which combines RDF technique with information retrieval, and provides users a friendly searching way, has also been a hot research area. But those approaches which directly search for results on RDF data graph utilizing a graph exploration algorithm, still cannot avoid the problem that by using keywords it is difficult to accurately express users’ search intention.An effective technique for keyword search over RDF databases is to incorporate an explicit interpretation phase that maps keywords in a keyword query to structured query. Because of the ambiguity of keyword queries, there are always too many interpretations for a unique keyword query. A typical solution is to generate top-k likeliest user-intentioned interpretations by using heuristics. However, it is more database-dependent, such as occurrence frequency of sub-graph pattern connecting keywords. Consequently, it often fails to capture user-dependent characteristics. Even the top-k interpretations cannot reflect accurately the users’real search intention.In this paper, we propose a context-aware approach for keyword search over RDF, which utilizes users’ query context to help interpreting keyword query. Firstly, we decide the relevance relationships among keywords from the last two queries. And then we interpret these keywords to obtain a set of query summary graphs, by actually dealing with the issue of mapping between keywords and class nodes in RDF schema graph. After analyzing the keywords’ relevancies and users’click-through information, we establish a group of policies to update the node weights and edge weights in each query context model (we also called it weighted context-aware query summary graph). We also provide approaches for evolving and updating such context model. At last, we design a graph exploration algorithm to search for query patterns. We rank these patterns according to our novel designed evaluation model. Extensive experimental results show that our proposed context-aware approach for keyword search over RDF can effectively improve the quality of searching results and their ranking. Also, our approach is efficient enough.