- From: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
- Date: Mon, 23 Feb 2015 09:34:33 +0000
- To: Michael Brunnbauer <brunni@netestate.de>
- Cc: "public-lod@w3.org" <public-lod@w3.org>
On 21 February 2015 at 20:38, Michael Brunnbauer <brunni@netestate.de> wrote: > I admit that what you sketch here is better than what I have sketched with > named graphs. But it seems to require a very sophisticated editor or a > very sophisticated user. Agreed that the editor needs to be more clever than three text fields with auto-complete. If a non-RDF user is to be making RDF, at a triple-level, then the editor must be guiding the user towards the basic Linked Data principles rather than be a glorified vim. :) > I was talking about an editor where the user can > add triples with arbitrary properties. .. and arbitrary properties/classes should include ones the user make up on the spot and relate it to existing properties/classes. Kingsley was talking about the sentences analogy, and I think we should keep this in mind. While you shouldn't normally write a whole book using your own terminology, it's very common to introduce at least SOME terminology that you specify or clarify for the scope of the text (e.g. the document, graph, dataset). It could be something as in a triple (a sentence), show something as simple as a little [+] button next to the property or class to specialize it and use this instead in that triple. (making it available for other triples) Perhaps Kingsley's approach have some mechanism for specializing or introducing new properties? > So I would prefer the solution with named graphs that can be implemented > easily. Yes, the additional information will be very vague without context > but it will at least be there. I wouldn't dismiss the named graphs-per-triple approach - although to me it sounds overkill (perhaps one named graph per transaction). However starting to encode relations between those named graphs to make them sequential (and thus encode some hidden semantic meaning) to me feels a bit like trickery - your user isn't really making RDF statements the way we commonly find it in Linked Data world - he's making some kind of custom graph-ordering-structure that is just represented in RDF. Using named graphs for provenance is however very much best practice - so if you just track the provenance of when triples were inserted and one of your views is to view the graph or statements about a subject in that order (present it as "Insertion order" or something), then you are good - you are not tricking the user anymore. You thus won't need to implement "Move up" buttons as you should not rewrite history. :) -- Stian Soiland-Reyes, eScience Lab School of Computer Science The University of Manchester http://soiland-reyes.com/stian/work/ http://orcid.org/0000-0001-9842-9718
Received on Monday, 23 February 2015 09:35:21 UTC