Spend Network Ground Truth Dataset For Entity Linking

This dataset provides manually annotated ground truth for entity linking over public procurement data from the Spend Network corpus, using Wikidata as the target knowledge graph. It includes: SN_input_table.csv: A structured table containing procurement records, where each row represents a contracting authority (buyer) with related address information (street, locality, postcode, country) SN_GT.csv: Ground truth annotations specifying which cells (identified by table, row, and column) are linked to Wikidata entities. The dataset also includes NIL entries for cases where no suitable entity could be found in Wikidata. This resource supports the evaluation of entity linking systems in a realistic setting where NIL detection is essential. It has been developed in the context of the enRichMyData project and is designed for semantic enrichment and benchmarking tasks involving tabular data.
Publisher name: Zenodo
Last updated: 2026-02-20T14:47:02Z