"Five years ago, I rolled into the semantic web world by accident. The Province of Limburg had the plan to digitize and publish their cultural heritage collections. The project manager had the visionary idea to do this using semantic web technologies. Amplexor was chosen to implement the plan because we had some experience with the technologies. We discovered however that the technology was not really mature enough to implement the whole vision: although we used semantic web technologies on the back-end, this was not shown in its full glory at the web interface (which is on the website http://www.erfgoedplus.be). Because the company behind the only software that was advanced enough to implement our vision went bankrupt, we couldn't ultimately realise this idea." 

"The basic selling point for the semantic web is the graph data model: instead of using a rigid scheme for your database, you serialise your dataset into triples of the form: subject - predicate - object, such as Alice - knows - Bob. Other semantic database creators do the same, and if you agreed upon the same identifiers for your data or if you are able to match them, you can connect your databases. This opens up a lot of possibilities, all because of the network effect: the bigger a network (in this case: the network of connected databases) is, the more useful it is. However, I find it difficult to talk to companies about the advantages of the semantic web: it's all about abstract concepts like serialising data, ontologies, automatically deriving new information, etc."

 

"What I do see is that some companies that want to integrate various data sources are investigating graph-based databases. A graph-based database is a good solution in a lot of use cases. Gone are the days when some analysts could sit around a table and design a data model that would be used for years. If the company then faces various market changes, it has to adapt the data model, which is a difficult and time-consuming process. A graph database is much more convenient to keep pace with our fast-changing world. Indeed, from a graph database to semantic web technologies is just one step further."

"In the last five years, this rather abstract chronicle of the semantic web has been supplemented by the Linked Data story, which is easier to grasp and more pragmatic. Tim Berners-Lee, the inventor of the web and the semantic web vision, introduced this idea: Linked Data is about using the World Wide Web to connect data that wasn't previously linked, or at least lowering the barriers to doing so. For creators of web applications, Linked Data is a big deal: instead of having to learn yet another API (application programming interface) for each source of data they want to use, they only have to learn one 'API': the Linked Data principles and underlying semantic web technologies. Later, Obama's administration launched the website data.gov that strives to make government more transparent by opening access to government data. This example was followed by the British government in data.gov.uk, led by a team under the supervision of Tim Berners-Lee himself. Thanks to these linked and open data stories, the semantic web suddenly became much more pragmatic: people could write their own mash-ups to visualise and explore government data."

"The semantic web still has a couple of big obstacles. An important problem is the quality of data. People are pretty good at evaluating the quality of documents on the 'classical' World Wide Web: they don't trust the information and read blindly, but take into account the reputation of the author, the author's affiliations, and so on. In the semantic web, there's a big difference: the data is analysed by computer programs. Unfortunately, computers are not that good at evaluating the quality of information."

"Another obstacle for the semantic web, and one that is more fundamental, is the open world assumption, which is typically made with an incomplete database such as the web: if you don't find a specific statement on the web, it doesn't mean that it's false, it just means that we don't know whether it is true or false. In contrast, if you want to implement semantic web technologies in a company, you'll find that most company databases are complete and hence you have to use the closed world assumption: everything you don't know to be true is false. Because of this mismatch, you need extra constraints to implement semantic web solutions in a company."

 

"And last but not least, let's not forget the end users. At the moment we don't have any frameworks that make it simple to build user interfaces for semantic web data. Almost all the techies in this domain are working on the back-end, such as algorithms for reasoning about data. This means that all semantic web projects now have to create ad-hoc interfaces for their end users. From a sales point of view this is an awkward situation: we can't easily build demos when we talk to a company for a potential semantic web project."

 

"Nevertheless, I think semantics will play a big role in many domains in the coming years, for example in terms of search functionality. Moreover, Drupal 7 will have built-in RDFa support. Hundreds of thousands of Drupal sites contain vast amounts of structured data, and thanks to RDFa this semantic data will be exposed in the HTML code generated by Drupal. For instance, if you create a new job posting on a Drupal website, semantic search engines would pick this up with all its properties, like salary, duration, and so on. Other services could then try to match supply and demand. This could turn out to be a killer app for the semantic web."