Benchmark Datasets for Entity Linking from Tabular Data
收藏资源简介:
📖 Benchmark Datasets for Entity Linking from Tabular Data (Version 2) This archive provides a benchmark suite for evaluating entity linking algorithms on structured tabular data.It is organised into two parts: • Challenge datasets (HTR1, HTR2): From the SemTab Table-to-KG Challenge, widely used in academic evaluations of table-to-KG alignment systems. Each is a dataset (a collection of many tables) provided with ground truth and candidate mappings.👉 Please also cite the SemTab Challenge when using these resources. • Real-world tables (Company, Movie, SN):• Company — one table constructed via SPARQL queries on Wikidata, with both Wikidata and Crunchbase ground truths.• Movie — one table constructed via SPARQL queries on Wikidata.• SN (Spend Network) — one procurement table from the enRichMyData (EMD) project, manually annotated and including NIL cases for mentions with no known Wikidata match. A shared top-level folder (mention_to_qid/) provides JSON files mapping surface mentions to candidate QIDs for these real-world tables. ⸻ 📂 Contents Each dataset or table includes: • One or more input CSV tables • Ground truth files mapping mentions/cells to Wikidata QIDs (or NIL) • Candidate mappings (mention_to_qid/*.json), sometimes multiple variants • Optional files such as column_classifications.json or cell_to_qid.json ⸻ 📝 Licensing • HTR1 & HTR2: CC BY 4.0 • Company & Movie: Derived from Wikidata (public domain; CC0 1.0) • SN: CC BY 4.0 (from the enRichMyData project) ⸻ 📌 Citation If you use these datasets, please cite: • This Zenodo record (Version 2):Avogadro, R., & Rauniyar, A. (2025). Benchmark Datasets for Entity Linking from Tabular Data (Version 2). Zenodo. https://doi.org/10.5281/zenodo.15888942 • The SemTab Challenge (for HTR1/HTR2):SemTab: Semantic Web Challenge on Tabular Data to Knowledge Graph Matching (Table-to-KG). (Cite the relevant SemTab overview paper for the year you reference.) • Wikidata: Data retrieved from Wikidata (public domain; CC0 1.0). • enRichMyData (for SN / Spend Network): Project resources from enRichMyData, licensed under CC BY 4.0.
📖 表格数据实体链接(Entity Linking)基准数据集(版本2) 本归档文件提供了一套面向结构化表格数据实体链接算法评估的基准套件。该套件分为两个部分: • 挑战数据集(HTR1、HTR2):源自SemTab表格到知识图谱(Table-to-KG)挑战赛,已广泛应用于学术领域的表格到知识图谱对齐系统评估。二者均为包含多表格的数据集,附带标注真值与候选映射关系。👉 使用此类资源时,请同时引用SemTab挑战赛相关文献。 • 真实世界表格(Company、Movie、SN): - Company:通过对Wikidata执行SPARQL查询(SPARQL)构建的单张表格,同时包含Wikidata与Crunchbase标注真值。 - Movie:通过对Wikidata执行SPARQL查询(SPARQL)构建的单张表格。 - SN(Spend Network,支出网络):源自enRichMyData(EMD)项目的一张采购表格,经人工标注,包含无对应Wikidata匹配项的空值(NIL)标注案例。 顶层共享文件夹mention_to_qid/ 提供了将表层提及映射至上述真实世界表格候选QID(Wikidata实体标识符)的JSON文件。 ⸻ 📂 数据集内容 每个数据集或表格包含以下文件: • 一个或多个输入CSV表格 • 标注真值文件:用于将提及/单元格映射至Wikidata QID(或NIL) • 候选映射文件(mention_to_qid/*.json),部分数据集存在多个变体版本 • 可选附属文件,如column_classifications.json 或 cell_to_qid.json ⸻ 📝 授权协议 • HTR1与HTR2:采用CC BY 4.0协议 • Company与Movie:数据源自Wikidata(公有领域;采用CC0 1.0协议) • SN:采用CC BY 4.0协议(源自enRichMyData项目) ⸻ 📌 引用规范 若使用本数据集,请引用以下文献: • 本Zenodo记录(版本2):Avogadro, R. 与 Rauniyar, A. (2025). 表格数据实体链接基准数据集(版本2). Zenodo. https://doi.org/10.5281/zenodo.15888942 • SemTab挑战赛(针对HTR1/HTR2):SemTab: 面向表格数据到知识图谱匹配的语义网挑战赛(Table-to-KG)。请引用您所参考年份对应的SemTab综述论文。 • Wikidata:数据源自Wikidata(公有领域;采用CC0 1.0协议)。 • enRichMyData(针对SN/Spend Network):源自enRichMyData项目的资源,采用CC BY 4.0协议授权。



