ALD/E Neuro-symbolic Query Benchmark: 33 Scientific Queries over Machine-Actionable ORKG Comparisons

This record contains the ALD/E Neuro-symbolic Query Dataset, a curated collection of 33 scientific queries (19 ALD, 14 ALE) defined over machine-actionable Open Research Knowledge Graph (ORKG) comparisons extracted from published review tables.

Each query bundle includes:
a natural-language question (brief + detailed forms),
the corresponding SPARQL gold-standard query,
CSV exports of the underlying ORKG comparison tables,
symbolic results (results_SPARQL.csv),
neural and symbolic-context-augmented results from 21 language-model systems,
machine-readable metadata linking to the source paper, DOI, ORKG comparison IDs, and query type.


The dataset supports research in NL→SPARQL translation, scientific table QA, symbolic vs neural vs neurosymbolic evaluation, and reproducible meta-analysis of ALD/E processes.It also includes domain-expert survey assessments of query clarity and result quality.

The resource is intended for materials scientists seeking FAIR, queryable ALD/E knowledge, and for AI researchers developing models that connect natural-language questions with graph-structured scientific evidence.