[Semantic Web]RDF(S) Storage in Relational Databases

读书笔记,  科学研究 Lee 发表于 2006/2/22 22:19:50 Although there are differences between RDF's graph-based model and the semi-structured model of XML, our work can benefit from the research on storing XML data in relational databases [Florescu and Kossman, 99]. Horizontal DB: Horizontal schema [Agrawal et al., 01] only need one “universal”table in the database. Every individual (instance) falls into one record in the table.While the data model is simple, there are some drawbacks within this approach: large number of columns; limits on property values; sparsity, etc. Vertical table: In [Alexaki et al., 01], it is also named the “Generic Representation”. This approach has a single table where each record corresponds to a RDF triple. However, this design means that any query has to search the whole database and queries that involve joins will be especially expensive. Horizontal class: This approach is similar to the horizontal database approach but there is a separate table for each class in the ontology. This essentially corresponds to the entity-relational approach frequently used when designing databases. Table per property: Yet another alternative is to assign a table to each property: PROPERTY_name(Subject,Object) In the database community, this approach is called the “decomposition storage model” [Agrawal et al., 01]. Like the vertical table approach, queries involving the implicit instances of a class can be particularly expensive. Hybrid approach: We adopted an approach that combines the property table approach with the horizontal class aproach (see Figure 1). It is similar to the “Specific Representation” in [Alexaki et al., 01]. According to this model, creating tables corresponds to the definition of classes or properties in ontology. The classes or properties’ ID should serve as the table names.                                                   ---- from DLDB: Extending Relational Databases to Support Semantic Web Queries

阅读全文(1649) | 回复(0) | 编辑 | 精华