FedSearch: Efficiently Combining Structured Queries and Full-Text Search in a SPARQL Federation

Combining structured queries with full-text search provides a powerful means to access distributed linked data. However, executing hybrid search queries in a federation of multiple data sources presents a number of challenges due to data source heterogeneity and lack of statistical data about keyword selectivity. To address these challenges, we present FedSearch — a novel hybrid query engine based on the SPARQL federation framework FedX. We extend the SPARQL algebra to incorporate keyword search clauses as first-class citizens and apply novel optimization techniques to improve the query processing efficiency while maintaining a meaningful ranking of results. By performing on-the-fly adaptation of the query execution plan and intelligent grouping of query clauses, we are able to reduce significantly the communication costs making our approach suitable for top-k hybrid search across multiple data sources. In experiments we demonstrate that our optimization techniques can lead to a substantial performance improvement, reducing the execution time of hybrid queries by more than an order of magnitude.