Hybrid Search with Multiple Full-Text Indexes over RDF Graphs
摘要
In this paper, we present an extension to the RDF4J SPARQL API called Filters that is providing hybrid full text search capabilities over RDF graphs. Unlike existing works in this area, our method allows creating and storing different full-text indexes. These indexes, depending on the chosen configuration, split the textual data on the spoken language or the RDF types they are associated with. This allows to index the data using different language analysers, as well as having higher query performance and smaller index sizes. As concrete application scenarios, we showcase two use cases on European projects, namely Kohesio, a platform that allows European citizens to find European Union (EU) funded projects in each region and the European Direct Contact Center (EDCC), a tool used by call center operators to answer 200k questions that are sent by citizens about EU. The code base is open source.