top of page

Dimensions of Interoperability in Agriculture - a report from the September ‘21 OGC Member Meeting

On September 15, 2021, at the OGC Member Meeting, the OGC Agriculture Domain Working Group (DWG) conducted a survey during their session that showed that more than half of the responders see the future of smart farming lies in decentralized cloud environments. The landscape of organisations participating in the Agriculture DWG session represent multi-level platforms and hundreds of heterogeneous data sources, interests, social habits, and legal regulations. In addition to this diversity, the high levels of fragmentation, weak cooperation, and traditional heritage that are characteristic of agriculture and livestock farming will, according to the McKinsey reports[1], stay in the tail of the digitized sectors. At the same time, agriculture is targeted as one of the Common European data spaces that shall allow for a free exchange concerning privacy and governance regulations.

The FAIR data principles that form the core OGC mission, and are realized by our open standards and the open data initiatives that use them, are ideally suited to addressing these challenges. Although the technical interoperability that OGC is focused on forms only one dimension of the problem known as the social-technical-semantic dimensional space, the possibility for stakeholders to connect disparate systems with each other is recognized as the key enabler for generating and exploiting required knowledge. Therefore, one of the key activities within the OGC Innovation Program is testing the standards, verifying their potential for adoption, and resolving interoperability issues with world-class experts. A particular challenge is that they have to be built vertically compatible with policies, user behaviors, and semantics. Cybele and similar projects are therefore an optimal frame to develop, test, and verify the required standards and technologies as they target bringing solutions to the market that address the many legal and business considerations.

[1] /figure/McKinsey-Global-Inst itute-industry-digitization-in dex-201513_fig1_336150205

These requirements align with OGC’s efforts in conceptual modeling and abstract specifications, but also in the development of the OGC Definition Server. One of the ideas behind OGC's work on the definition server is to support standardization by providing a tool to organize and manage standards-related ontologies. One of the initial use-cases was a reference host of the taxonomies that could be used to improve the quality of the observations gathered by the social sciences.

Currently, the OGC Definition Server is multi-purpose and built around the triple store database engine RDF4j with customized VocPrez front-end and a number of ingestion, validation, and entailment scripts. The validation and entailment are exploiting the potential of the SHACL Shapes definitions for TBox models of the entities, and narrows down the variations in conceptual models as to ground base common features of the instances and infer any not-explicitly defined second-order relationships (predicates). That meaningfully simplifies the A-Box definition of the instances. For example, reverse predicates (like next vs previous), other ontology relationships (dependency as skos:narrow or dcat:partOf) can be unambiguously inferred without explicit definitions or verified for inconsistencies.

In practice, that is a step forward in defining the bridge between conceptual and logical models. The concepts can be expressed as instances of various ontological classes and interpreted within various contexts, while their minimum definition can be translated into entities, relationships, and properties.

Currently, it is hosting several types of definitions covering:

Register of OGC bodies, assets, and its modules

Ontological common semantic models (e.g. Cybele, Demeter)

Dictionaries of subject domains (e.g. PipelineML Codelists)

Content of the core is incrementally growing as the data produced during several projects, like Cybele, Demeter, e-shape, GEOE3, or CLINT, is added. For the agriculture domain, it provides a ​​Common Semantic Model allowing mapping to the various standards/ontologies including FIWARE, Saref4Agri, INSPIRE, FOODIE, ADAPT, AGROVOC, EU standards, ISO standards. Containing also the model of the standards like Topic 20: Observations and Measurements, and the dependencies of SensorThings API, it shall be possible to define the reference model(s)’ profile, encodings, and code lists. Such precise definition is currently unachievable without specifying target encoding schema, which has to be produced manually with not-insubstantial effort. Ideally, precise models of the APIs shall reach the level known from plug-and-play hardware devices that quickly become accessible for nearly everybody.

What may be even more important in the future is linking the reference model to the external ontologies. Doing so can greatly improve the quality of the knowledge produced based on the collected data. Verifiability of the research outcomes and explainable AI are just two examples where a clear log of inferences and unambiguous semantic compatibility of the data will play a key role. Contextualization of the information is not a new problem: the Frame concept was proposed in 1974 (M. Minsky), but today’s increasing amount of data, coupled with the increasing capabilities of artificial inferring has brought the problem to the forefront of common data spaces’ integration.

Note: The article is based on the results of the Cybele and other researches presented in Agriculture presented during the Agriculture Domain Working Group during 120th OGC Member Meeting Sep’21 by OGC, Plan4All, PSNC.



bottom of page