"There's our SPARQL endpoint." Or "Just view the page in Tabulator." I have lost count of the number of times that either of these have been the only response to an innocent request to see what some new piece of semantic wizardry can do. For a developer seeking to integrate one semantics-rich data set with another, SPARQL may very well be the tool for the job. And for someone (probably a developer, again) who wants to track the way that data is pulled together to build a page, Tabulator has a lot going for it. But as a shop window for the power of semantics? As a demonstration of what's possible? Seriously, is it possible to pick worse ways to show off to the world?
In January's episode of the Semantic Link, we were joined by serial entrepreneur Nova Spivack (perhaps best known to readers as the Founder and CEO of Twine) for a discussion about the importance of delivering a good user experience. In the time available, we only scratched the surface, and I'm sure it's a topic to which we'll return.
One of the issues we touched upon was the challenge of delivering rich — semantically powered — functionality without overloading the user. A lot of the discussion revolved around one of Nova's latest startups, Bottlenose, but the issues are far more widely applicable.
Displaying results meaningfully, and enabling complex interaction, has been part and parcel of the Semantic Web toolbook for a very long time. MIT's SIMILE Project, for example, began early in the last decade and produced a number of tools that made it relatively straightforward for users to interact with timelines and other widgets. Although the project itself has ended, much of the thinking lives on in demonstrators from SIMILE alumni; David Huynh's Parallax tool, or Zepheira's Viewshare project for the Library of Congress, to pick just two.
But looking beyond SIMILE and its immediate descendants, examples of accessible interfaces to applications built upon the Semantic Web Stack become harder to find. Where are they hiding? Surely they must exist? Does the Semantic Web Stack not lend itself to the types of data that require nice interfaces? Do Semantic Web developers lack the ability to build a UI for normal people? Do tools built on the Semantic Web Stack make it hard to feed data to nicer user interface tools and conventions? Do either developers, clients or end users lack the conceptual frameworks that enable them to understand — and craft — the requirements for a UI that's about more than filling in text boxes?
There are, of course, plenty of examples of semantically rich web interfaces that look good, work well, and succeed in burying the nasty bits away from easily frightened users. But, it seems, (almost?) all of these make no use of the Semantic Web Stack with its OWL, its RDF, and all the rest. TripIt, Siri, Bottlenose, Zeebox. 'Semantic', useful, powerful, usable, pretty, reasonably intuitive. But 'semantic' and 'on the web' rather than formal Semantic Web.
What, if anything, does this tell us? And should we be concerned? Where are the rich, powerful and pretty equivalents from the formal Semantic Web world? Are they possible, but just not interesting enough for anyone to have got around to building? Or is there a more worrying technical or conceptual problem that we need to tackle?
The Data Web is dead in the water if its shop window is a SPARQL endpoint.