With the explosive growth of Linked Data, abundant RDF data have beenpublished on the Web. More and more researchers focus on how to access dataefficiently and accurately. Since traditional Information Retrieval (IR) technologiesare no longer suit for the retrieval on Linked Data, it becomes difficult for ordinaryusers to retrieve the data efficiently and accurately. We need a new retrieval methodfor accessing Linked Data.This paper presents a method of doing semantic keyword search on Linked Data.We propose two distributed inverted index schemes based on MapReduce frameworkand Bigtable Data Model, one of which is built from Linked Data and the other fromthe ontology. And as a necessary part of the ontology index, an ontology encodingscheme is also be proposed. Based on the index schemes, we design an improvedranking algorithm named OntRank by introducing a semantic factor into the BM25Franking model. The experiments evaluate the performance of distributed invertedindex from the indexing time, response time, and stress test, and use informationretrieval comprehensive benchmark to evaluate the accuracy of the OntRankalgorithm.Our work can be regarded as an extension to the traditional IR probabilisticretrieval model. The method not only retains the convenience of the keyword search,but also improves the retrieval accuracy for semantic keyword search. This work hasprovided a new way for the research of semantic keyword search. |