All articles

Linked-data UI’s: a response to David Karger

I wasn’t at this year’s ESWC conference, so I missed David Karger‘s keynote talk on A semantic web for end users. I’m sure it was a great talk, and I hope there will be a video version on the ESWC web site someday soon.

Reading the tweet stream, one of the issues that David was discussing was the lack of live end-user applications – or even screenshots of working apps – that accompany papers on semantic web and linked data research. Still less the kinds of robust user evaluation that UX people regard as the norm in other fields. I agree, and it’s a shame that, notwithstanding the good efforts of a core group, interest in semantic web UI’s has languished somewhat.

Part of the problem may be systemic. It seems to me that there’s a gulf between purely generic presentations of linked data, which take no account of the user’s goals and context of use, and presentations that are built around good UX principles of knowing your user and helping them solve their problems. In a well-designed task focussed UI, the fact that it’s linked data ‘under the hood’ is invisible to the user, who just wants to get their task completed effectively and efficiently.

We have worked on this problem at Epimorphics. The Environment Agency’s linked data on bathing water quality (note to American readers: bathing water in this case refers to coastal and inland water where people swim or go boating, not bathwater!) has a purely generic presentation through the linked data API. However, to help people find beaches of interest to them we also wrapped that API in a data explorer application and a widget designer to let web site owners embed local water quality data on their own sites. The data explorer has a greater task focus than the generic linked-data pages, but there’s still a slightly uncomfortable balance between meeting the needs of, say, a family wanting simply to know if the water at the beach near their holiday destination is clean at the moment, and a researcher or data journalist looking to build a story out of a deeper dive into the current and historical detail. The widget is simpler and even more task focussed, but then we have to ask: is it a linked data application, or an application which happens to use linked data?

In a more recent project with the Environment Agency, we’ve been looking at inland water quality more broadly under the aegis of integrated water catchment management. Here the value of linked data is more appreciable, as part of the strategic objective is for one authoritative agency (the EA) to provide a spine of reference data which can be used to integrate data collection and processing activities by a range of disparate and autonomous organizations. However, as the underlying data model has become richer, so the challenges of presenting a compelling UI to the end-user have become greater. We don’t have a live version of the catchment management proof-of-concept system yet, as it’s still in evaluation. However, there is a presentation about it on SlideShare, which I’ll be giving at SemTech San Francisco this year.