SALT-KG: A Benchmark for Semantics-Aware Learning on Enterprise Tables

SALT-KG is a benchmark dataset for semantics-aware learning on enterprise tables, presented at the EurIPS'25 Table Representation Workshop. The dataset is hosted by SAP and was last updated in May 2026. It includes metadata for an Operational Business Knowledge Graph.

Use Cases

• Benchmarking semantic learning models based on enterprise table data.
• Training models for knowledge graph construction from structured business tables.
• Evaluating table representation techniques for downstream tasks like entity linking or relation extraction.

Strengths

• Dataset is associated with a peer-reviewed paper presented at a NeurIPS workshop.
• Includes metadata for an Operational Business Knowledge Graph.
• Last updated on 2026-05-10, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

SAP

Collection Method

Likely constructed from enterprise tables for benchmarking purposes.

Freshness

Last updated 2026-05-10 11:03:49.