Interoperability & Linked Data

Why interoperability matters

Conceptual diagram of the POSO ontology relating a positioning system, its deployment, tracked entities and the algorithms and technologies used
Core concepts of the POSO ontology — a positioning system, its deployment, the entities it tracks and the technologies it relies on. From POSO: A Generic Positioning System Ontology (ISWC 2022).

Proprietary data formats create silos. Every new integration becomes a bespoke interface, every vendor change becomes a migration project, and the knowledge about what the data means stays locked in the heads of the people who built the system. Interoperability inverts this: the meaning travels with the data, so any software that understands it can take over.

This is also the direction European policy is moving in. The EU's ambition for a Digital Single Market and initiatives such as Gaia-X and the International Data Spaces all depend on shared vocabularies and machine-readable semantics. Building on open standards today means you are not retrofitting compliance later.

Interoperability also has a direct commercial benefit: a system that can describe itself is a system that third parties can build on. That is the difference between a closed deployment and a platform.

What I do

Ontology & vocabulary design

Domain-specific ontologies in OWL and RDFS that capture how your organisation actually models its domain — reusing established vocabularies wherever they exist, rather than inventing another silo.

  • Alignment with existing standards (SSN/SOSA, GeoSPARQL, schema.org, DCAT)
  • W3C best practices for publishing and versioning vocabularies
  • SHACL shapes for validation and contract-style data quality checks
  • Human-readable documentation generated from the ontology itself

Linked data integration

Turning existing relational, CSV, JSON or sensor data into a queryable knowledge graph without forcing a big-bang migration of your source systems.

  • RML/R2RML mappings and transformation pipelines
  • SPARQL endpoints, federated queries and query optimisation
  • JSON-LD APIs so existing web clients need no changes
  • Triple store selection, deployment and performance tuning

Decentralised & sovereign data

Architectures where the user — not the application — holds the data, based on the W3C Solid specification and data space principles.

  • Solid pods as interoperable personal data vaults
  • Access control, consent and audit trails
  • GDPR-aligned data minimisation by design
  • Data space and Gaia-X compatible sharing patterns

Discoverability & service discovery

Making systems findable and self-describing, so that clients can discover what a service offers without prior coordination or a proprietary registry.

  • Self-describing endpoints and semantic service descriptions
  • Beacon- and network-based discovery of nearby services
  • Capability negotiation between independent systems
  • FAIR (Findable, Accessible, Interoperable, Reusable) data practices

Key concepts

A short glossary of the terms used on this page, based on my PhD research and on talks I give to government and industry.

Interoperability

The ability of independent systems to access, read and understand each other's data, so software built by one party can work with data produced by another. In practice it answers one question: can another piece of software run on the same data? It has three layers:

1. Accessible

The data can be found (for example through a catalogue) and reached (for example through an API), with the right permissions.

A data consumer finds the data of a producer through a catalogue and API, with the right permissions

2. Readable

Both sides use the same open, standardised format (JSON, XML, CSV), so no proprietary driver is needed. Layers 1 and 2 are called syntactic interoperability.

A producer and a consumer exchange data in the same open standardised format

3. Understandable

Both sides refer to the same shared concept with an unambiguous definition, so the meaning survives the exchange. This is semantic interoperability.

A producer and a consumer refer to the same shared concept with an unambiguous definition

A fourth layer, processing interoperability (also called pragmatic or behavioural), describes how data was or will be processed, for example which algorithm produced a value and with what accuracy, so another system can take over the process and not just consume its output.

Semantic interoperability

Two systems can exchange a perfectly valid JSON file and still misunderstand each other. Does "date": "03/04/2026" mean 3 April or 4 March? Is a "customer" the person who pays or the person who uses the product? Syntactic interoperability gets the bytes across; semantic interoperability makes sure the meaning arrives too. It works by pointing at a shared definition instead of describing the concept in your own words. When two systems use the same definition, they mean the same thing.

Application-independent data

Data that is described, stored and accessible separately from the software that produced it, so another application can run on the same data. The data becomes the long-lived asset and applications become interchangeable pieces of software on top of it.

  • Own the data, not the app: it lives somewhere you control, not only in the vendor's product.
  • Describe the data, not the app: the meaning of each field is machine-readable, so the original code is not needed to interpret it.
  • Open interfaces: any application with permission can read, and where allowed, write.
  • Replace without migrating: a new application is pointed at the same data.

Data silo and vendor lock-in

A data silo is data that can only be accessed or interpreted through the one application that holds it; vendor lock-in is the resulting dependency on that vendor. Silos grow because every application has its own data model and every connection needs its own translation. With n applications the number of translations grows towards n × (n − 1), and every model change can break several of them.

Three applications (CRM, invoicing and support) each have their own data model and transformations to every other application
Each application translates to every other one. This does not scale.

Linked data and RDF

Linked data is the technology that makes application-independent data practical. Data is published as RDF statements of the form subject – predicate – object. Every concept is named with a web address (a URI) that resolves to its definition, so data is self-describing and can be connected to other data over the web. Standardised web technologies (URIs, RDF and HTTP) provide both syntactic interoperability (HTTP) and semantic interoperability (shared vocabularies). The price is some overhead: data that explains itself is larger than a compact format both sides already know, a trade-off worth making when systems do not know each other in advance.

A subject connected to an object by a predicate
The basic building block of RDF: a subject, a predicate and an object.
The concept man is labelled "man"@en and has the broader concept human
Statements combine into a graph. Here, the concept man is narrower than human.

Ontology and vocabulary

An ontology (or vocabulary) is a shared, machine-readable definition of the concepts in a domain and how they relate. It is what the “shared concept” in semantic interoperability points to. Good practice is to reuse established ones, such as SOSA/SSN for sensor observations, GeoSPARQL for geography, DCAT for data catalogues and schema.org for general use, and only define new terms where none exist.

SPARQL

The query language for RDF, comparable to SQL. Its strength is that a query can follow links across several sources, so one question can be answered from data that lives in different places.

Data shapes (SHACL)

A vocabulary says what the words mean; a data shape says how the data must be structured. SHACL describes constraints such as required properties, types and value ranges (a latitude must lie between −90 and 90), so data can be validated before another system relies on it. Shapes act as a contract between producer and consumer.

Personal data vault (Solid)

Storage that the data owner controls, independent of the applications that use it. In the W3C Solid specification this is a Pod: the owner chooses where it is hosted and decides which application may read or write which part of it, using a WebID for identity and access control per resource. It is the most direct form of application-independent data: applications visit your data, not the other way around.

FAIR data

A set of principles for data that is Findable, Accessible, Interoperable and Reusable, by machines as well as by people. They describe the same goal as this page from the viewpoint of data management and are increasingly expected in European data policy.

Research behind the practice

The vocabularies below were designed and published as part of doctoral research at the WISE lab of the Vrije Universiteit Brussel. They are open, freely reusable, and actively maintained — and they are the reference implementations behind the consultancy work above.

The Positioning System Ontology extends SSN/SOSA with the concepts needed to describe a positioning system, the environment it is deployed in, and the algorithms and techniques used to produce a position — not just the resulting coordinates.

A semantic Bluetooth Low Energy beacon specification, backwards compatible with iBeacon, Eddystone and AltBeacon. A SemBeacon broadcasts a link to linked data describing itself, its environment and the positioning systems available nearby — no proprietary beacon database required.

A fiducial marker ontology that classifies visual markers, their spatial pose, and the detection and pose-estimation procedures applied to them — enabling augmented reality applications from different vendors to interpret the same markers in a shared space.

The open-source hybrid positioning system that puts these vocabularies to work: a modular TypeScript framework where positioning techniques are nodes in a processing graph, with built-in RDF serialisation and Solid persistence.

Selected publications

POSO ontology concept diagram
POSO: A Generic Positioning System Ontology
ISWC 2022
Mobile application showing position data retrieved from a Solid pod
A Solid-based Architecture for Decentralised Interoperable Location Data
IPIN 2022
SemBeacon application showing detected beacons on an indoor map
SemBeacon: A Semantic Proximity Beacon Solution
ACM IoT 2023
Augmented reality objects anchored to printed fiducial markers
FidMark: A Fiducial Marker Ontology for Semantically Describing Visual Markers
ESWC 2024

The full list — including the doctoral dissertation Interoperable and Discoverable Indoor Positioning Systems — is available on maximvdw.be.