Abstract
We will here present a framework for digitally documenting materials knowledge addressing all levels of semantic documentation, including cataloguing as well as structural, contextual and semantic documentation. It ensures that the documented data becomes FAIR with a special emphasis of making it meaningful, linked and actionable. The documentation itself is stored in a knowledge base, while the documented resources (e.g. datasets) may be stored elsewhere. All types of resources (datasets, materials, processes, samples, instruments, models, researchers, etc…) can be documented and related to each other using the Elementary Multiperspective Material Ontology (EMMO) and its domain ontologies. It ensures consistency of data documentation with materials science knowledge. Datasets are furthermore documented using the DCAT-AP vocabulary (making them accessible and useful for DCAT-aware data services) and preferably also with formalised data models mapped to the ontologies (to document their structure and achieve interoperability at a numerical level).
The framework is developed with the perspective of the data producer focusing on making the documentation process as seamless and simple as possible, while ensuring that all aspects of the FAIR principles are addressed. The data producer is only exposed to well-defined keywords that may either refer to literal values or other resources in the knowledge base. Behind the scenes, the keywords are mapped to ontological data properties (literals) or object properties (resources) using a JSON-LD context. The set of available keywords can be extended with a custom JSON-LD context. To keep the required input for the data documentation to a minimum, the framework reuses the mappings and data models to prepopulate all input fields that can be inferred from the documented resource. Complex (potentially nested) inferences and transformations are achieved with the help of so-called mapping functions. They are functions typically implemented in Python, who’s input and output are defined from the ontologies. They are documented like any other resource and applied transparently by the framework whenever transformations are needed between different literal representations.
To document a series of similar resources, the framework offers a spreadsheet-like interface. This allows the data producer to prepare the documentation using a familiar tool like Excel and commit the documentation of all the resources in one go.
The European Union‘s Horizon 2020 and Horizon Europe research and innovation programs are acknowledged for funding via OntoTrans (862136), OpenModel (953167), MatCHMaker (101091687) and PINK (101137809). The Research Council of Norway is acknowledged for funding via SFI PhysMet (309584).