What semantic interoperability means

Interoperability is the ability of public bodies to work together by sharing information through their business processes and the data their systems exchange. The European Interoperability Framework (EIF) describes it in layers, and the semantic layer is the one that makes sure both sides understand the data in the same way.¹ Without it, two systems can connect perfectly and still pass each other information that is wrong in practice, because a word such as "address" or "household" means one thing in one agency and something else in the next.

The layers of interoperability

The current EIF, published by the European Commission in 2017 as an annex to Communication COM(2017) 134, sets out four layers: legal, organisational, semantic and technical. Integrated public service governance cuts across all four, and interoperability governance sits in the background.¹ The guide to the EIF covers the rest of the framework.

The layers depend on each other. A data-sharing agreement is no use if the law doesn't allow the exchange, and a fast, secure connection doesn't help if the data that flows through it is misread at the other end.

What the semantic layer adds

The EIF sums up the aim as "what is sent is what is understood". It splits the semantic layer into two parts. The syntactic part describes the exact format of the information, its grammar and structure. The semantic part covers the meaning of data elements and the relationships between them, through vocabularies and schemas that every party reads the same way.¹

Shared formats are the easier half. Two agencies can both send well-formed files with a field called address, and validation will pass on both sides. Whether that field holds a postal address, a property, a usual residence or a correspondence address is a question about meaning, and no file format answers it. Semantic interoperability is the work of agreeing that meaning, writing it down and attaching it to the data.

What breaks without shared meaning

Addresses

Addresses look simple until two systems try to match them. One record says "Flat 2, 14 High Street", another says "14a High Street", and a third holds only the building. In Great Britain the government's answer is a shared identifier: systems, services and applications that store or publish data containing property and street information must use the Unique Property Reference Number (UPRN) and the Unique Street Reference Number (USRN).⁴

A UPRN identifies every addressable location, including things with no postal address, such as an electrical substation or a public toilet. A block of flats can have one grid reference but a separate UPRN for each flat.⁴ The identifier doesn't replace the address. It is added as an extra field, so that records held by different bodies can be linked to the same place.⁴

Households

"Household" is harder, because the disagreement is about who counts. The harmonised definition for UK official statistics is one person living alone, or a group of people, not necessarily related, at the same address who share cooking facilities and share a living room, sitting room or dining area.⁵ It was used for the 2011 Census and in Office for National Statistics (ONS) social surveys from 2011; earlier censuses used a slightly different test, based on the only or main residence and sharing at least one meal a day or the living accommodation.⁵

A benefits service, a housing register or a council tax system may need a different unit altogether, such as a couple and their children, or the people named on a tenancy. None of these is wrong. The trouble starts when a figure labelled "household size" moves from one system to another without saying which definition produced it.

How vocabularies, code lists and reference data help

The EIF treats agreements on reference data, in the form of taxonomies, controlled vocabularies, thesauri, code lists and reusable data models, as key prerequisites for semantic interoperability. It recommends an information management strategy set at the highest possible level of an organisation, with priority for metadata, master data and reference data.¹

• Core vocabularies. The Commission's Core Vocabularies are simplified, reusable and extensible data models for basic things such as a person, a legal entity, a location and a public service. They can be a starting point for new systems, the basis for exchanges between existing ones, a way to combine data from different sources and a common export format for base registries such as business registers.² The Core Location Vocabulary, for example, describes a location as an address, a geographic name or a geometry, and is aligned with the data specifications of INSPIRE, the EU's Infrastructure for Spatial Information in the European Community.¹²
• Code lists and authority tables. A code list fixes the values a field may hold and what each one means. The EU Publications Office maintains authority tables, also called named authority lists, to harmonise codes and labels across systems and make exchanges between institutions easier.³
• Reference data that says what it is not. The authority table for types of administrative territorial unit warns that its units must not be confused with the statistical regions of the Nomenclature of Territorial Units for Statistics (NUTS), which are defined by population, or with the areas courts cover, even where they coincide in some countries.³ Three bodies can each mean something different by "region".

The guide to ontologies and core vocabularies explains how these models are written down and published. Shared identifiers and definitions matter most in the authoritative registers that other services rely on, covered in base registries and data exchange.

Why the semantic layer is the hard one

The EIF itself says this layer poses significant challenges, because Member States differ in language, culture, law and administration.¹ Many definitions are set in legislation, so an agency cannot simply adopt someone else's meaning; changing it can be a legal question as much as a data one. Older systems add to the problem, because their meanings often live in code and field names rather than in documentation.

Opinions differ on how far to standardise. One approach is a single shared model that everyone adopts; another lets each body keep its own model and publish mappings to a common core. The EIF recommends supporting sector-specific and cross-sector communities that create open information specifications and share them on national and European platforms.¹ In practice most programmes mix the two approaches, and both have costs: agreeing a definition takes time, and one written to suit everyone can end up suiting no one well.

The payoff is greatest where services cross organisational or national lines. The EU's once-only evidence exchange, for example, only works if an authority in one country knows what a document issued in another country actually proves.