Comments on EmotionML Working Draft 29 October 2009


Reading through the working draft I noticed some parallels to other work and would like to suggest alignment.

First of all seing that you define a metadata format it would seem natural to do things in RDF but I can see how a strict XML format is more suitable for the use-cases and for combining this format with other formats as pointed out in section 5.2. And I'm sure you considered RDF anyway since you reference it at one point. :-)
Nonetheless it seems that you could still use ontology formats (OWL, SKOS) and URIs for the definition and use of vocabularies, e.g. for the <category> element.

Regarding the <metadata> element: The definition says "This element can be used to annotate arbitrary metadata." - I guess this should be something along the lines of "to annotate something *with* arbitrary metadata"?
And yes, for the content using RDF would make a lot of sense as you say, especially Dublin Core properties would work well here. But then maybe you should define what the "scope" or the "origin resource" is that those properties get attached to when you want to interpret this as RDF. If the <metadata> is a child of <emotionml> then it should probably be the URI of the document. But what about <metadata> within <emotion>? Should <emotion>s have URIs? Or should such metadata be interpreted as being attached to RDF "blank nodes" (which get defined through the attributes of the <emotion> element)?
If you would restrict the contents of <metadata> a bit more you could even define a simple conversion to RDF via GRDDL [1].
The elements within <metadata> in section 5.1.1 are not namespace-scoped btw. so I wonder if that's maybe a bit confusing for the reader - they're not part of EmotionML, right?

Looking at the <link> element I think this should follow the same style as <link> elements in other formats such as HTML and Atom: @uri would be @href and @role would be @rel. Otherwise it might confuse users who are used to those other formats.
Maybe since you define a <link> element you could allow @xml:base as well? (This would then also influence the "origin resource" of <metadata>.)

And finally, seeing how reference time in <emotion> and <link> I wonder if using the work of the Media Fragments WG [2] would be suitable?



Received on Saturday, 31 October 2009 13:35:08 UTC