OKP4 Ontology principles
What is an ontology ?
In computer science, ontology is a formal and structured representation of the concepts, relationships, and properties of a particular domain. An ontology generally comprises the following basic elements: concepts, relationships, properties, axioms, and instances. These can be graphically represented by the simplified equation shown below.
- Concepts: represent the main formalized elements of the domain.
- Relationships: represent links between concepts.
- Properties: represent specific attributes or characteristics of the concepts themselves.
- Axioms: represent logical statements or rules that define relationships between concepts, properties, and instances, ensuring the consistency and coherence of the knowledge represented within the ontology..
- Instances: the concrete instances of concepts representing objects in the application domain. In OKP4, instances are used to represent all the resources of the dataverse.
Some examples of ontology:
An ontology of sheep and goat (source : OKP4):
An ontology of water resources (source : OKP4 from SAREF extension for water) :
Why ontologies ?
For OKP4, ontology is essential as it enables the description of shared knowledge. Participants can better understand and interpret the exchanged information, even if they come from different backgrounds.
This ontology allows us to achieve:
- Standardization of terminology: standardized terminology is used for concepts and relationships in a given domain, clarifying and avoiding misunderstandings between participants.
- Structuring of data: data is structured in a coherent and organized way, making it easier to access, process, and analyze.
- Interoperability of systems and tools: a well-designed ontology enables interoperability between systems and tools, facilitating the sharing of knowledge among different stakeholders.
- Improved data research and analysis by accurately describing concepts and relationships in a particular domain.
The knowledge representation language chosen for OKP4 is RDF Schema and Web Ontology Language on top of the framework Resource Description Framework.
A formal model for the OKP4 blockchain
This ontology describes and defines the different forms of vocabularies used in the OKP4 protocol in a standard and well designed format. The aim is to model a semantic network of all the entities (Data Spaces, data, services, processing workflows) by semantically characterizing what they are and the relationships they maintain between them. Thus, the ontology provides a complete living understanding and knowledge of the datasets within a Data Space, their transformation (by the services), as well as the governance rules that apply (data sharing, consents, policy rules).
Ontology at the heart of the blockchain
Ontology is at the heart of the OKP4 protocol as much of the information is encoded and stored as an ontology on-chain in the blockchain transactions. This means that (almost) all the semantics of the transactions submitted to the blockchain are expressed through this ontology - for instance the creation of a dataspace, the execution of a service, the description of a dataset, etc.
The OKP4 ontology
The OKP4 protocol orchestrates the various resources of the Dataverse (datasets and services) using different blockchain elements such as smart contracts, logic modules, and ontology. All these elements allow for fine management of dataset and service workflows for knowledge creation within a Data Space with personalized governance. As seen previously, the ontology must stand for the different concepts of the protocol, their relationships, and their properties.
The following diagram depicts the introduced concepts and their relationship with the already existing concepts of the ontology.
Class and properties
The following concepts and properties are found within the OKP4 ontology:
This refers to the data contained within a dataset.
This is a dataset made available by a user on the protocol.
This is the description of a given dataset in metadata form.
A Data Space groups resources.
A decentralized identifier URI. A URI that identifies a subject in a decentralized system and is managed independently of any centralized registry.
The information data about something (i.e. data about the data). This something can be a Dataset, a Service, a Dataspace, or any other entity that can be described. Metadata is an abstract concept which is refined in Metadata Profiles used to provide a formal specification that defines the set of metadata elements, their semantics, and their syntax to be used in a particular context or application. The OKP4 protocol proposes several profiles at the core of the ontology, such as GeneralMetadata for describing services or datasets.
Services or datasets, a resource belongs to a Data Space.
- hasIdentifier A service consumes a resource and produces data.
This ontology in OWL language is written as follows:
rdfs:comment "Define a data"
rdfs:comment "Define a data"
rdfs:label "Dataset General Metadata"
rdfs:comment "Define a data"
With all these concepts, their properties, and their relationships, we can create the OKP4 ontology and explain the workings of the OKP4 protocol in a structured and formalized way. This ontology can be expressed in different formats, more or less understandable by humans or machines. It can be expressed in French or English, RDF, OWL, JSON-LD, N-Triples, Notation3 RDF/XML, Turtle, etc.
Last released version of OKP4 ontology documentation is available here: https://ontology.okp4.space.
- There's no one correct way to model a domain and a trade-off must be found between the meaning given to ontology, its expressiveness, its extensibility and its usage.
- The OKP4 ontology is not frozen. It is built step by step in an iterative process (see next section), and some decisions made here may be changed later.
- It should be understood that OWL modeling is different from UML modeling (or more simply of the Oriented Object interpretation that one would be tempted to make). As such, the following readings are relevant:
- OWL being a logical description language, some deductions can be made by an OWL reasoner. However, as far as possible, it will be best to make explicit what could be deduced by an OWL reasoner.