Publication of version 1 of the Garance website (disseminating the knowledge graph generated from the reference datasets of the Archives nationales de France)

[CLAVAUD Florence]

Hello,

 

I am pleased to announce the launch of version 1 of the Garance web application, available online at https://rdf.archives-nationales.culture.gouv.fr/garance  (use this link to directly access the website in English: https://rdf.archives-nationales.culture.gouv.fr/garance/pages/en/.

Garance (a French acronym for Graph of the Archives nationales de France, for Research, Access, and Navigation of Enriched Knowledge) is the experimental website for the dissemination, through knowledge graphs, of the reference datasets of the Archives nationales de France (AnF).

These reference datasets consist of indexing vocabularies and authority records, collaboratively produced by AnF archivists. To date, they describe approximately 18,400 agents, 54,000 places, and 3,400 concepts, which serve as contextual entities for the records preserved by the institution. This is an incomplete and imperfect, yet unparalleled knowledge repository, given the scope of AnF’s work and the expertise required to describe these entities. These datasets are of interest to a wide range of users and are already being used in various research projects.

However, the AnF’s information system, which manages the non-RDF source of these datasets, does not provide users with a way to access them (except for record creators or accumulators), nor does it allow them to exploit their inherent nature as a directed graph.

The AnF Lab developed in 2021, and it has maintained since then on GitHub (see here), a semantic version of these institutional reference datasets. It is currently finalizing a version 2.0 of these datasets, compliant with the latest version of the Records in Contexts - Ontology (RiC-O 1.1) and with SKOS. These datasets have thus been restructured, standardized, and transformed into a graph; they have also been enriched through various automatic or semi-automatic processes. Notable enrichments include the creation of a vocabulary of place types and the enhancement of place reference datasets.

While waiting for a new information system to be designed and implemented, including optimized management and access functionalities for this reference datasets, the goal of the Garance project is to provide all users with a web solution for accessing this RDF data, thereby making it compliant with the FAIR principles, while exploring new ways of navigating, searching, and consulting the data to inform the design of the future IS.

The Garance website is being developed by the AnF Lab in partnership with the company Sparna. The functionalities of Garance version 1 include:

- A menu providing access to an alphabetical list of entity names (for places and agents) and to some diagrams providing general statistical information;

- A quick search function within the content of the descriptions;

- Access to web pages for viewing all available information for each entity (one page per agent, one page per place, one page per controlled vocabulary);

- Buttons for downloading data in RDF/XML format (and CSV for vocabularies or EAC-CPF for records on archival creators);

- Access to these pages via the RDF URIs of these entities (content negotiation);

- A SPARQL data access service for third-party applications;

- A direct query interface using SPARQL, including pre-recorded queries;

- A page about the project and the technical architecture of Garance, whose main components are a static website generated from RDF data, on the one hand, and the open source QLever triplestore, on the other;

- A contact page.

Garance's interface is available in French and English.

Version 2 will be released in autumn 2026; it will include new data and new features.

Best regards,

Florence Clavaud

Head of the Lab, Archives nationales de France

Chair of ICA/EGAD, lead of RiC-O development team

---

Merci de nous aider à préserver l'environnement en n'imprimant ce courriel et les documents joints que si nécessaire.