In this blog, we present a summary description of five data models that were chosen for internal testing of the AI4PublicPolicy Semantic Interoperability Toolkit. These data models were selected based on their capability to represent a diverse range of information. It is important to note, that these can often work together and complement each other. Rarely will there be a ready-made model that adequately fulfills all requirements or does everything well. These Data Models may be repurposed to suit specific needs, functioning at times as modules that can be coupled with each other to represent more complex data.

1. Smart Data Models

The Smart Data Models aims to provide a wide range of data models for multiple domains. From Smart cities to Smart water and energy, having a common way to model the data from different domains helps ensure that there is a technical common ground, which then enables open innovation and procurement. In the following figure, the Smart Data Models home page is presented.

2. UnitsNet

UnitsNet provides a robust library with definitions for physical quantities and their units. This data model does not immediately correlate to Smart Data Models but can complement existing works by ensuring the that said models correctly reference the unit or units that they use. Moreover, this library specifies conversions between units (for instance, from miles to meters or vice-versa).

3. Semantic Sensor Network Ontology

The Semantic Sensor Network Ontology is a W3C Recommendation which includes two main specifications: the SSN (Semantic Sensor Network) and SOSA (Sensor, Observation, Sample and Actuator). SOSA is the core of SSN and its main aim is to provide support for a wider audience and areas of application that use Semantic Web ontologies, while also being a minimal interoperability fall-back. The figure below presents the SOSA and SSN ontologies and their vertical and horizontal modules.

4. QUDT

Originally developed for the NASA Exploration Initiatives Ontology Models (NexIOM) project as part of the Constellation Program Initiative at the ARC (AMEAS Research Center), QUDT is a public charity non-profit organization that makes ontologies, derived models and vocabularies. QUDT is a collection of several ontologies that are linked to each other as shown in the following figure.

QUDT’s main utility is to provide unit and quantity definitions ensuring interoperability across domains and records. Thus, it is used in SSN and can also be applied to other data models where unit definitions are limited such as in the case of Smart Data Models complementing them. The use cases for QUDT are unit conversion between single and complex types, dimensional analysis of equations, finding equivalent units in different systems of units and finding equivalent quantity kinds in different systems of quantities.

5. SAREF

The Smart Applications Reference (SAREF) ontology specifies common concepts in smart applications, the relationships between them and the axioms to constrain the usage of the described concepts and relationships. It provides a set of building blocks that can be added or removed, according to the identified requirements, leading to an ontology that is reconfigurable.

Data Models analysis and comparison

The different data models possess differing qualities. While the Smart Data Models were created specifically for Smart contexts, they lack flexibility for contexts that may have more unique needs. This is not the case for SSN. It is a much more flexible tool that can be applied to a variety of different scenarios. However, SSN may be more difficult to understand and use when compared to Smart Data Models.

UnitsNet does not capture dynamic entities and their properties, but it does provide a comprehensive and actively maintained library of physical quantities and units, serving a role similar to QUDT. This is crucial for ensuring interoperability because merely having the value from an observation is insufficient; one also needs to know the associated unit. Smart Data Models lack sufficient specifications in this regard. Therefore, it can be used in a manner similar to how QUDT is used in SSN.

SAREF presents itself as a more complete and repurposable ontology for multiple domains. It already considers the question of unit types as opposed to Smart Data Models. When compared to the SSN, they are very similar being both modular, of a more general character and developed with a vision for use in “Smart” contexts.

As such, Smart Data Models presents itself as a ready-made solution for specific domains where someone has already developed a data model to use “out of the box”. SSN and SAREF represent ontologies of a more general character, enabling interoperability across domains but with the caveat of requiring more user knowledge for their use.

QUDT and UnitsNET can offer support for unit definitions in cases where the model does not provide them. These narrower data models provide more precise descriptions and can be used either as an alternative or as an extension to other models.