HybIdx: Indexes for Processing Hybrid Graph Patterns Over Text-Rich Data Graphs

Many databases today are text-rich in that they not only capture structured but also unstructured data. In this work, we propose a full-text extension to SPARQL capable of expressing different types of hybrid search queries over text-rich RDF graphs. We study existing indexing solutions to arrive at the conclusions that database extensions are less efficient than native solutions, and the most efficient indexes are limited w.r.t. the types of hybrid search queries they can support, i.e. entity queries. We propose an indexing solution call HybIdx, which is both efficient and versatile in terms of query type support, i.e. it supports full SPARQL graph patterns where keywords can appear in any position (relational queries). Experiments suggest that it can outperform the second best approach by up to three orders of magnitude for complex queries.