Linked Data has a lot of potential, but in absence of a thorough business case, it remains one of those promising technologies with little impact. Here are some thoughts about the business case.
Before going into that, however, it is worth noting that @Noud Hooyman represents the principal of several central government registries related to land administration. The first truly large-scale Linked Data project in the Netherlands aims at making the information in several of these registries publicly available in RDF, in other words, as Linked Open Data. The idea is that information about buildings and parcels are projected on a map, enabling you to interact with the data and explore different kinds of relevant information about these, from different sources. For instance, you could see that a building is inhabited, and then click through to the piece of legislation that defines these terms. Since the data, their interconnections and their semantics are machine-readable, interesting applications using this information are expected to emerge. So, the challenge for a business case for Linked Data comes from an important source.
The revolutionary idea of Linked Data is that the semantics of data is made available alongside the data itself in a machine-readable way. Suppose we make a dump of the data from some application. In a certain column in a certain table we see dates. But what does it∏ mean? Date of birth? Date of burial? Date of… what? To interpret these dates, we may look at the column header. With any luck, we find a descriptive name, in the worst case nothing or something unintelligible. But even if the column header is descriptive enough to give us some notion of what the dates mean, this information is not precise enough for reliable reuse. You have to understand the application that generated the data, or at least the data model it uses. The idea of Linked Data is that the semantics of the data in such cases can be looked up on the Web, using a very simple, standardized method that also works for machines.
Transitioning to Linked Data therefore frees up data from the systems they are generated in, making them independent, accessible, reusable and combinable. If you know the semantics of data residing in different data sets, you can combine them seamlessly, without expensive translation protocols created by programmers who understand the respective data models: that work can now be automated. Big Data requires clear and explicit definitions of information elements. Linked Data enables you and your computer to trace a piece of data to its definition in a dictionary, a glossary or thesaurus. Machines do not (yet) understand the textual dictionary definition, but they can now ascertain whether two things mean the same or not. This is essential when combining data from different sources. Linked Data enables Big Data to be applied at Web scale.
What is the business case in that? This would be a misguided question. You can’t ask for the business case of the Web itself either. A business case only exists for specific projects delivering specific results. Now that is precisely where the catch is. The project referred to earlier results in publication of massive data sets in RDF. This costs a lot of tax payer money, so the question is: for what?
To justify this, we need projects that create value based on these data sets. One type of business case is based on replacing existing mechanisms for data exchange by Linked Data technologies. @Ria van Rijn and @Arjen Santema wrote a paper some time ago arguing that this would drive down the cost of data exchange in the context of authentic base registries in the Netherlands with tens of millions of Euros per year. While the theory behind this is totally sound, it remains a theoretical argument until real projects make this happen.
Another type of business case would be presented by projects that create new value instead of making existing mechanisms of value creation cheaper. As in the early days of the Web, it is clear that this will happen, but we cannot predict exactly when and how. An interesting demonstration of what is possible now that the land registration data become available in RDF has been developed by some people in the project — mainly @Laurens Rietveld and @Wouter Beek of Triply, the amazing Linked Data spin-off of the Vrije Universiteit of Amsterdam, and @John Walker of Semaku. They created a web page showing the most densely populated municipalities in the Netherlands on a map. The shapes representing the municipalities come from the BAG and the BRT, two of the land registry data sets hosted at Kadaster. The population figures are found in Wikidata, the native Linked Open Data sibling of Wikipedia and DBPedia. Note that the data are retrieved and combined real time. And that the project took hours to complete, not months. In the old days, this would have costed a fortune. (Note that the web page is experimental and of a temporary nature: sorry if at some point the link is broken.)
Again, this is not a “real” project with a “real” business case, it is just a demonstration of what is possible. The problem with the semantic Web and Linked Data is that you need a critical mass of data on the Web before the network effect kicks in and uptake follows a hockey stick-shaped growth curve. For the mass to grow critical, you need enough organizations publishing useful data in RDF. Which, of course, they are only willing to do if there is a real business case for it. GOTO 1.
Thanks to visionary decision makers in the area of land registration in the Netherlands — and Noud Hooyman certainly ranks among them — we can now look forward to a wealth of land registry data becoming available as Linked Data. This is a significant step towards reaching critical mass. What these and other decision makers need is creation of real value based on this, to justify further investments in the same direction. I am sure that the thriving community of the Linked Data crowd in the Netherlands will soon come up with something. We cannot predict exactly when and how, but it will happen.