A MultikeyRank Model Based on Ontology for Large-Scale Semantic Data

In order to provide users with intelligent retrieval services over large-scale semantic data, this paper proposes a MultikeyRank model. ORDPATHs is adopted to encode ontology classes defined in TBox and an inverted index is constructed according to the structured characteristics and semantic association of RDF data. A new query mode with one primary keyword and several auxiliary keywords is designed. To reflect the user's perference objectively, the membership degree for ontology classes corresponding to the primary keyword is calculated based on the evidence theory and thus the MultikeyRank algorithm is formulated by extending the BM25F model accordingly. The proposed model was implemented in the selfdeveloped distributed large-scale RDF data server "Jingwei" and experimental results show that compared with BM25F, the evaluation indexes for P@5, P@10, P@15 and MAP are improved by 27.6%, 24.3%, 18.5% and 3.7%, respectively.