Het2onto Benchmark
Table of Contents
1. Our Benchmarks
2. Benchmark Expectation
2.1. Components
As in obda-benchmark.org, our benchmark is composed of the following components:
- An source database
- A set of mappings
- An ontology for the target schema
- A set of queries on the ontology
2.2. Characteristics
We want that your benchmark have the following characteristics:
- the source schema should contain relations with different attribute numbers like common relational database,
- the target schema have to represent RDF graph schema, or DL schema if we don't support variable for class or property in mapping head,
- the mapping head should have mostly star shape BGP,
- some mapping head should have small chained star shape BGPs with existential variables, like: \[\mathrm{Author}(x) \rightarrow \triple{x}{\type}{\mathrm{Person}}, \triple{x}{\mathrm{writes}}{y}, \triple{y}{\type}{\mathrm{Book}}\]
- the target schema constraints should represent an RDFS schema.
Questions:
- Do we want to set manually the source or the target schema ?
- Do we want to set manually the target schema constraints i.e. ontology ?
- Is it better to use already known OBDA system as a starting point or only use synthetic integration system ?
2.3. Related Benchmarks
There are several benchmarks having all the required components, but not having the required characteristics. The following list contains some of them made for OBDA scenarios:
- Texas Benchmark has one source relation about Product. It enables 6 integration scenarios, each having is own mappings, ontology and queries. The mappings interest is in triples of form \(\triple{x}{\type}{\class}\), where each class \(\class\) has a different selectivity according to the source table and the ontology.
- BSBM
- NPD Benchmark was done in Ontop project. It has an OWL 2 QL ontology. Initial dataset can be scale using the scaler tool Vig.
- IMDB Benchmark
- iBench is a benchmark generator creating mappings according to source and target schema and link between them.
- We have explored other RIS based on BSBM.