Best Practices Part 2

Agriculture currently faces numerous challenges on a global level: There is a need to produce and deliver enough food for a still growing world population in a sustainable manner without depleting natural resources. To fulfill that task, it relies on knowledge from a number of scientific disciplines including for example biology, geosciences, economic and social sciences, and engineering domains like machinery construction. Many of the questions currently arising can only be addressed by combining data from these different disciplines coming from distributed, heterogenous sources. From the information management perspective, agriculture is thus not a single domain that could be covered by a domain data model, but rather a domain of domains. Data integration is therefore a challenge. Recent developments around flexibly combinable RDF descriptions of (meta)data, provide solution approaches but are currently not yet consolidated and widespread enough. However, within the European GAIA-X cloud infrastructure as well as Research Data Management Initiatives like the German National Research Data Infrastructure (NFDI) such approaches are used and tested also within an agricultural context. Data catalogues using combinations of different vocabularies (Dublin Core, DCAT, schema.org and more) are currently being established. The contribution highlights a few aspects of reuse practices applied like for example subproperty hierarchies for identifiers, using SKOS concepts in property ranges, how to deal flexibly with depth of data provided etc. That way, a higher degree of interlinkage between (meta)data systems can be achieved that especially helps cross-domain applications like the ones prevalent in agriculture.

How to identify lessons learned and actively use them instead of lossing them to the lessons learned database.

A core metadata model allows for harmonizing diverse data and data sources, offering end users a unified way to access their data. At the IMF, this core metadata modeling work focuses on capturing core metadata elements across the IMF, covering diverse business cases, and building on enterprise vocabularies, following established market standards (Dublin Core / DCAT/ SDMX Data Structures etc.). In this presentation, we will focus on how combining a core metadata model and the promise of a unified discovery interface and search help manage IMF data in the short and medium term.

Research data management (RDM) includes people with different needs, specific scientific contexts, and diverse requirements. The description is a big challenge in the domain of RDM. Metadata plays an essential role, allowing the inclusion of essential information for the interpretation of data, and enhances the reuse of data and its preservation. The establishment of metadata models can facilitate the process of description and contribute to an improvement in the quality of metadata. When we talk about image data, the task is even more difficult, as there are no explicit recommendations to guide image management. In this presentation, we present a proposal for a metadata model for image description. To validate the model, we followed an experiment of data description, where eleven participants described images from their research projects, using a metadata model proposed. The experiment show that participants do not have formal practices for describing their imagery data. Yet, they provided valuable contributions and recommendations to the final definition of a metadata model for image description, to date nonexistent. We also developed controlled vocabularies for some descriptors. These vocabularies aim to improve the image description process, facilitate metadata model interpretation, and reduce the time and effort devoted to data description. We hope that the intended audience of this presentation will be, above all, researchers who use and produce data in image format, data curators, and managers of data deposit and publication platforms, so that they become aware of the importance of convenient image management in science.

Linked Open Data (LOD) is touted as a key technology to enable data sharing and interoperability. But how exactly must this technology be implemented to deliver the promise of creating FAIR data, Linked Data that is findable, accessible, interoperable and reusable? We present our work on interoperable semantic data models and how we implement them in practice to create a pan-institutional knowledge graph of cultural collections for digital research in art history.