Medieval Charters Knowledge Graph (MCKG)

Medieval Charters Knowledge Graph (MCKG)

This repository contains all supplemental materials -including the datasets, code, and validation artefacts- supporting the paper:

MCKG: A Community-Driven Knowledge Graph on Medieval Charters

Submitted to the Semantic Web Journal.

The Medieval Charters Knowledge Graph (MCKG) is a provenance-aware knowledge graph constructed from the AMSPO medieval charter corpus, modeling persons, institutions, events, properties, and genealogical relations described in medieval Spanish charters.

The live Wikibase instance is publicly available at Wikibase cloud. RDF datasets (Turtle format) used to populate the MCKG are available in the datasets/ directory.

Corpus description
Source corpus: AMSPO medieval charter corpus

Document type: Medieval legal charters (e.g., donations, property transactions, institutional acts)

Geographical scope: Asturias, Spain

Temporal coverage: 14th century

Language: Medieval Spanish

The corpus contains documentary records describing legal, economic, and social relationships between individuals and institutions. Each document entity in the MCKG includes the archival identifier used in the AMSPO corpus (e.g., AMSPO, FSV, nº XXXX), enabling users to locate the corresponding charter in the scholarly transcription cited in the original paper (Felpeto Cueva J. El archivo de un artesano del siglo XIV: el orfebre Alfonso Fernández de Oviedo. doctoral thesis, University of Oviedo, 2023. URL https://digibuo.uniovi.es/dspace/handle/10651/71402. Accepted: 2024-02-16T11:28:22Z.).

Knowledge Graph Scope
The MCKG provides a structured semantic representation of documentary information extracted from medieval Spanish charters (AMSPO corpus). It models:

Legal and socio-economic acts recorded in charters (e.g., donations, property transfers), including their temporal context and role-based participation of agents.

Natural persons and institutions appearing in the documents.

Explicit kinship and institutional affiliations asserted in the text.

Properties involved in legal transactions and ownership transitions.

Provenance distinguishing domain expert annotations from community-supported contributions.

The data model combines the event-centric semantics of CIDOC-CRM with selected properties aligned to Wikidata, implemented within a Wikibase environment and validated through Shape Expressions (ShEx).

Dataset Versioning and Statistics

Version

Current version: v1.1.0

Release date: 2025-05-13

Last update: 2026-03-11

Statistics (v1.1.0)

15542 statements.

2211 entities.

7196 qualifying statements.

18810 reference statements.

128 documents processed.

Version history
v1.1.0 (current release)

Added mentions entity helper property to document entities to improve navigation between documents and referenced entities.

Added missing is documented in statements for some location entities.

Added explicit reference to the scholarly source representing the AMSPO corpus.

Expanded and clarified repository and Wiki documentation.

Expanded and reorganized SPARQL example queries.

Improved visibility of SPARQL queries and RDF/Turtle datasets.

Updated dataset statistics.

v1.0.0

Initial public release of the MCKG dataset.

Reproducibility
This repository provides all artefacts required to:

Inspect the RDF datasets derived from the AMSPO corpus.

Reproduce the RDF-to-Wikibase ingestion workflow.

Examine validation constraints (ShEx / EntitySchemas).

Re-run SPARQL-based statistical analyses.

The RDF datasets provided here correspond to the data ingested into the public Wikibase instance (version v1.1.0). The live MCKG instance and the repository together ensure reproducibility of the population pipeline described in the paper.

Repository Structure
All resources are bundled in mckg.zip and organized as follows:

datasets contains RDF datasets (Turtle format) produced by applying our pipeline to the AMSPO corpus:

Community Dataset includes the RDF files generated from community contributions via semi-automatic analysis.

Domain Expert Dataset includes the RDF files created from manual expert annotations.

code contains Jupyter Notebooks implementing key pipeline components:

CSVGenerator performs RDF-to-CSV conversion for Wikibase ingestion.

Medieval NER with Roberta performs domain-specific NER to facilitate data extraction by community contributors.

Shexer extracts Shape Expressions from the RDF datasets for validation and EntitySchema construction.

WikibaseIntegration loads into the MCKG core entities and properties of the data model, as well as the processed datasets in CSV format.

csv contains the csv files used in WikibaseIntegration. Those with the nomenclature wikibase_import_X_dataset are an output of CSVGenerator.

shapes stores both Shape Expressions extracted by sheXer and the resulting EntitySchemas, available as well in the MCKG for validation.

sparql contains the SPARQL queries used for generating dataset statistics and for the analytical examples discussed in the paper. These queries are also available as executable examples in the MCKG Query Service. The queries are organized by research theme, including:

Charter and transaction analysis.

External knowledge integration.

Genealogical reconstruction.

Linguistic and onomastic analysis.

Prosopographic and sociographic analysis.

Provenance-aware queries.

Spatial analysis.

Statistics.

NER contains RoBERTa NER outputs, used as support for community contributions.

This README is included in both the ZIP file and repository root for clarity.

Intended audience
The MCKG is intended for:

Historians working with medieval charter corpora.

Digital humanities researchers.

Semantic Web researchers.

Contact
The MCKG was developed by WESO research group.

Responsible researchers:

Jorge Álvarez-Fidalgo, Universidad de Oviedo

Enrique Rodriguez-Martin, Universidad de Oviedo

Jose Emilio Labra-Gayo, Universidad de Oviedo

For inquiries, collaboration proposals, or technical issues, contact [e-mail].