The exploration of large, unknown RDF data sets is difficult even for users who are familiar with Semantic Web technologies as, e.g., the SPARQL query language. The concept of faceted navigation offers a user-friendly exploration method through filters that are chosen such that no empty result sets occur. Computing such filters is resource intensive, especially for large data sets, and may cause considerable delays in the user interaction. One possibility for improving the performance is the generation of indices for partial solutions. In this paper, we propose and evaluate indices in form of the Bitmap Triple (BT) data structure, generated over the Header-Dictionary-Triples (HDT) RDF compression format. We show that the resulting indices can be utilized to efficiently compute the required exploratory operations for data sets with up to 150 million triples. In the experiments, the BT indices exhibit a stable performance and outperform other deployed approaches in four out of five compared operations.
展开▼
