Restructuring And Application Toward Chinese Knowledge Graph

With the rapid development of Semantic Web, more and more RDF repositories, such as Linking Open Data(LOD), are available on the web.Generally, there are two services provided for exploring those RDF repositories, one is the keyword lookup, and the other is the SPARQL endpoint.Most users choose the lookup service, and millions of web logs have been recorded. Although, users expect to submit more expressive queries than keyword lookup, the complexity of SPARQL undoubtedly scared users away. What’s more, while the RDF model is the standard for Web data interchange, for graph computing the property graph model gains increasing attentions along within the emerging No SQL databases. A number of property graph stores have been developed and used in many real-world applications. However, there is still no widely accepted performance benchmark for property graph stores.Therefore, how to utilize the existing RDF repositories has become an open task in Semantic Web. Firstly, we propose a method of SPARQL query recommendation for exploring RDF repositories. By analyzing web logs of the lookup service, our method extracts the user access patterns,which will be used to recommend SPARQL queries. We implement our method based on Zhishi.me, a Chinese RDF repository with about 150 million triples as well as over one-year web logs.Then we also propose a method to build property graph benchmarks based on existing RDF repositories. By converting RDF graphs to property graphs, our method enables the property graph benchmark to run with real data rather than synthetic data. Furthermore, by analyzing the query log of the RDF repositories, a set of graph data access patterns can be extracted and used for benchmark query generation, to mimic real-world graph query workloads. We implement the proposed method based on Zhishi.me, and use the benchmark to evaluate popular property graph stores including Neo4 j and Titan.