Linked Data Innovation for a Smarter Government

Today is Prinsjesdag, the day that the Miljoenennota, the Dutch government’s yearly national budget is published. This budget determines how government funds will be allocated to achieve societal goals in the coming year. The Miljoenennota initiates the budget implementation cycle which consists of various phases: spending, adjusting, auditing and reporting.

During this cycle, the Dutch Ministry of Finance must pull together a multitude of reports, or information products. These products are created in several data workflows that combine detailed financial information originating from various government sources. These data workflows involve specific business logic which requires detailed domain knowledge to be applied precisely. In the past, many steps in this workflow consisted of manual data copying operations and implicit knowledge available only in the heads of experts. As a result, the workflow could not be automatically repeated or validated.

Triply and the Ministry of Finance are now working on a proof-of-concept to implement increasingly larger parts of this workflow using linked data. This allows data to be interchanged, combined, processed, and validated in increasingly repeatable and automated ways.

Adding meaning to data

When data is stored in traditional databases, the meaning of the data is encoded within the database’s schema. However, these dataset schemas cannot be easily shared across databases. Since traditional databases use a dedicated schema, the meaning of the data remains local to that database. Where complex organizations use hundreds of databases, their information is effectively dispersed over a large number of isolated data silos.

Triply’s approach to data integration is to use a global schema, or ontology, to encode the meaning of data in a way that can be shared among databases.

With the help of TriplyDB, the Ministry of Finance is creating a standardized ontology of national budget concepts. For example, the ontology distinguishes commitments to future payments from payment transactions. In traditional databases, the distinction between commitments and transactions must be enforced by the database user. This requires knowledgeable users, and even knowledgeable users can accidentally exchange commitments for transactions, resulting in incorrect reports. By using an ontology, the system is aware that the commitment amount is conceptually different from the transaction amount and can warn the user.

By storing information in line with this national budget ontology, the Ministry can automatically integrate financial information from different government agencies without alignment mistakes. Secondly, since the ontology encodes the meaning of the key financial concepts and the constraints they must adhere to, the validity of the financial information can be verified automatically. Lastly, the financial information can be combined with other information from different domains altogether, offering more insights to both the public as well as the financial professional. This opens up interesting applications such as label-based budgeting: for instance, ‘green budgeting’ or ‘gender budgeting’.

Back to the source 

Storing financial information in linked data allows information to be traced throughout the entire budgetary workflow. For example, a number that is published in a cell of a table in the Miljoenennota can be traced back to the very calculation that produced it. The calculation itself can be traced back to the source data on which it was based. Within an organization this workflow can be traced across datasets and across processing tools. The figure 1 shows a diagrammatic overview of some of these relationships.

The high-level workflow relationships are standardized in the international PROV [1] linked data ontology. Because linked data uses universal identifiers, it can even be traced across organizations. When organizational boundaries are crossed this must of course occur in a secure way, using secure web-based protocols.

Figure 1: Derivation of the Miljoenennota information product using linked data.

Linked data can be used in many domains

While the Dutch government is adopting linked data in several areas, the potential of linked data innovation is by no means limited to government organizations. Because linked data is a domain-independent technology, it can be applied to every domain for which a formal ontology can be created. Linked data is used by many other Dutch government agencies. For instance, the Cadastre (Dutch: Kadaster) is using geospatial ontologies to disseminate its key registers of property ownership and property value (see Kadaster Labs for examples). The Dutch Digital Heritage Network (NDE) uses cultural heritage ontologies to integrate data collections maintained by hundreds of different heritage organizations. This results in one unified Dutch digital heritage collection, through the use of linked data.

Finally, linked data is also making its way into the research community.  Another example is the Cooperation Databank: a database that contains linked data descriptions of thousands of sociology papers. This annotated literature collection allows large-scale research analyses to be carried out, with the potential for new and interesting academic findings.

The European Commission and the G20 have agreed to improve academic data publication practices on a global scale. This initiative is known under the acronym FAIR: research data should be Findable, Accessible, Interoperable and Reusable. As with the government examples, scientific data can be more easily found, accessed, interchanged, and reused when it is stored as linked data. The Common Lab Research Infrastructure for the Arts and Humanities (Clariah) is using TriplyDB to publish research data in a FAIR way. This allows humanities scholars to reuse each other’s data in innovative ways.

Conclusion

Linked data is becoming more mainstream and is now commonly applied in large production systems. The Netherlands Standardisation Forum recommends (RDF, OWL) and requires (SKOS) its use. The need to apply linked data techniques is also increasing: information workflows are becoming more complex, traditional data integration approaches have reached their limits, and workflows increasingly include statistical approaches like machine learning. Linked data provides the semantic standards that allow organizations to meaningfully represent and interchange all aspects of their complicated information workflows.

This also means that there is a new challenge for data education programs: to increase the semantic literacy of domain experts and data scientists. In the near future scientists, analysis, journalists and civil servants will use linked data ― either explicitly or implicitly ― as part of their daily workflow.

[1] UvA professor Paul Groth was instrumental in the standardization effort of PROV, see https://www.w3.org/TR/prov-overview/