[Datasets] Select, curate and standardize benchmark datasets (BSBM, LUBM, Bowlogna, DBpedia)

[remiceres]

Summary

Select, download, generate, and standardize official benchmark datasets across varying sizes and topologies (synthetic and real-world) for systematic evaluation.

Objectives and Technical Scope

1. Dataset Selection and Sizing

• Synthetic benchmarks:
• Berlin SPARQL Benchmark (BSBM): e-commerce domain, mix of simple and complex queries.
• Lehigh University Benchmark (LUBM): university domain, ontology with reasoning potential.
• BowlognaBench: student/university relationships.
• Real-world benchmarks:
• Curated DBpedia samples across standard distributions.
• Establish standardized scale tiers: Small (~50k triples), Medium (~1M triples), Large (~10M+ triples).

2. Storage and Preprocessing

• Provide automated download/generation scripts and verify checksums.
• Normalize serialization formats (N-Triples, Turtle) to ensure parsing consistency across different evaluated stores.

Acceptance Criteria

[ ] Dataset generation and download scripts are automated and committed to the repository.
[ ] Datasets across small, medium, and large scales are formatted, validated, and documented.