We are interested whether the data model (relational vs. RDF) has an effect on downstream question answering quality in the semantic parsing setting (Text2SQL vs. Text2SPARQL).

This project is about:

• Converting existing relational benchmarks to RDF; possible candidate benchmarks are
  • Spider 2.0 (https://spider2-sql.github.io/)
  • BIRD SQL (https://bird-bench.github.io/)
• Evaluating QA methods on the converted RDF data and checking, whether and in which cases the RDF data model helps