The work presented in this paper was carried out as part of the MemoMines project, which aims to provide scientific tools and methods to help preserve and promote the mining heritage of the Hauts-de-France region (France). A major social challenge in the field of cultural heritage, particularly mining heritage, is linked to the collection, analysis, publication and promotion of the collective history and memory of those involved in the field, both oral and written. In this sense, formalising information on cultural heritage represents a real challenge due to the diversity and incompleteness of the data. Furthermore, this data is heterogeneous and can be found in various sources, both online and offline: databases, libraries, museums, press kits, stakeholder expertise, etc. This diversity of resources poses numerous challenges, such as data documentation, representation, integration and interoperability within a single knowledge base.

To address these issues, we propose a methodology for building a knowledge base, also known as a mining heritage domain ontology (coal mining) limited to the departments of Nord and Pas-de-Calais in France and covering a period from the start of mining (around 1750) to the present day, including, for example, the rehabilitation of former production sites and mining-related habitats. In this regard, we wish to structure and formalise knowledge characterising the stakeholders, places, temporal entities and themes present in the analysed oral and written documentary corpus, and the ontology produced should then enable us to facilitate access to it.

The proposed methodology consists of three stages, the first being the formalisation of named entities (places, stakeholders, temporal entities) and a structured vocabulary describing the mining domain in an initial ontology of the mining domain. Temporal entities correspond to calendar references (dates and years). For named entities such as stakeholders and places, we draw on resources produced by Mission Bassin Minier, a project partner that manages the UNESCO label on behalf of the state. For the domain vocabulary, we rely on the ThesoMines thesaurus defined as part of the project (Daloz and Chaudiron, 2019), validated by experts in the field and available in its 1.0 version with open access on the OpenTheso platform provided on the servers of the Très Grande Infrastructure de Recherche HumaNum. The thesaurus, defined according to the SKOS (Simple Knowledge Organisation System) standard, consists of 558 descriptors and 764 non-descriptors, and is structured into four domains (culture, education, economy, natural environment, history) and five facets (activity, built environment, deposit, material, operation) in order to cover both tangible and intangible heritage, as defined by UNESCO.

With regard to formalisms, most attempts to resolve semantic interoperability issues focus on standardisation, or even normalisation, of common formalisms such as FRBR (Le Boeuf, 2013), FRBRoo (Doerr et al., 2008), CIDOC CRM (Doerr, 2003), etc. In light of the state of the art and our initial experiments (Kergosien et al., 2019), we have chosen to formalise the ontology produced in CIDOC CRM, a reference conceptual model designed to model cultural heritage domains. We will use concrete examples to detail the choices made in formalising the knowledge derived from the selected resources (thesaurus, list of places, etc.).

The second step consists of describing the corpus of oral and written documents based on the initial ontology produced. This second step corresponds to the instantiation (or populating) of the ontological model. Our work here is part of the open data policy that has gained momentum over the last decade. In line with the Law for a Digital Republic (7 October 2016) strengthening the open data policy, the Ministry of Culture has committed to a policy of openness and sharing of public data, in particular to enable citizens to participate in improving data quality. As a result, the National Archives, in close collaboration with the interministerial service of the Archives of France, has just published a new regime for the reuse of the public information it holds, whether in the form of inventories or archival collections. These are documents that are freely available to all and over which third parties do not hold intellectual property rights. The information must be freely communicable within the meaning of Article L.213-1 of the Heritage Code and must not have been communicated by authorisation or derogation. It is also within this framework that the National Archives have published several freely reusable corpora on the Ministry's open data platform: data.culturecommunication.gouv.fr. In 2019, the Ministry of Culture continued its programme by launching the POP open heritage platform, a new tool for managing the ministry's cultural data, comprising two parts: (1) a production platform, which allows the submission of metadata (images, photographs, descriptive or regulatory information); (2) a platform for disseminating this data, accessible to the general public. As part of the project, we are working with our partner INA to integrate our domain ontology into the OKAPI platform dedicated to the indexing, analysis and publication of audiovisual data corpora. In order to explain this second stage of instantiation, we will draw on a detailed analysis of an initial interview conducted with a former miner to demonstrate both the value of our approach and the difficulties encountered in relation to indexing, particularly the choice of the level of granularity of description used.

The third step consists of enriching the initial ontological model produced with new knowledge gleaned from interviews with former miners (new terms, new entities named, etc.), then validating the resulting knowledge base with experts in the field. We will illustrate this final step with various examples drawn from our work.

The ontology produced will soon be made available in OWL CIDOC CRM format under a CC-BY licence, in order to make it accessible to as many people as possible. Our approach can be generalised to other endangered heritage sites and, in this sense, associated documentation incorporating the entity identification schemes that we are able to mark up from a document will also be made available to anyone wishing to use CIDOC CRM in a specific case related to heritage enhancement.