Domain Specific Data Retrieval on the Semantic Web

The Web content increasingly consists of structured domain specific data published in the Linked Open Data (LOD) cloud. Data collections in this cloud are by definition from different domains and indexed with domain specific ontologies and schemas. Such data requires retrieval methods that are effective for domain specific collections annotated with semantic structure. Unlike previous research, we introduce a retrieval framework based on the well known vector space model of information retrieval to fully support retrieval of Semantic Web data described in the Resource Description Framework (RDF) language. We propose an indexing structure, a ranking method, and a way to incorporate reasoning and query expansion in the framework. We evaluate the approach in ad-hoc retrieval using two domain specific data collections. Compared to a baseline, where no reasoning or query expansion is used, experimental results show up to 76% improvement when an optimal combination of reasoning and query expansion is used.