Re: Types of annotation

Hi Ralph, Phil.

Phil and I have been talking about this, and I think his
point also has further ramifications about what an annotation as
described in my simple annotations schema should look like, which I
hadn't noticed before.

Here's part of the W3 annotation schema:

[]----type---[annotation]
  ----annotates----[http://whatever]
  ----body----"some literal text"

and part of the one I've been using

[]----type---[annotation]
  ----annotates----[http://whatever]
  ----title----"Whatever"
  ----description----"some literal text"

In my simple annotations system, I regard the title and description
as referring to the annotated document, i.e. something like 'the
annotator says that the title of http://whatever is "Whatever"'. But
actually (I belatedly realize) this structure implies that the title is
the title of the annotation. 

What we're trying to do is to associate either:

some data or
some metadata

with a document. By which I mean, there are two cases, one where you're
associating two documents in a particular fashion (like typed href - see
for example Dan Brickley's nodes and arcs document 
http://www.w3.org/1999/11/11-WWWProposal/). This may be essentially the  
same as associating metadata with a document - as Phil and I have been
doing by creating anonymous nodes and content that only exists in a
database, in the sense that there is some description of the annotated
document, and this forms the annotating document, like the value of the
body property in the W3 schema.

However, I was trying to be more specific than saying this document
annotates this one - I was trying to say that document A says that
the appropriate audience of document B is adults only, for
example. As Phil says, we're trying to associate detailed metadata with
document A, the annotation, _about_ document B.

to do this, I think we need a structure like this:

[]----type---[annotation]
  ----annotates----[http://whatever]
  ----assertsThat---[]
		 ---subject----[http://whatever]
		 ---predicate--audience
		 ---object----"adults only"

rather than use my simpler example above, which means something
different altogether.

The W3C annotations schema avoids this problem by not specifying the
content of the annotation in a marked-up format, which is fair enough,
but I know Phil needs something like a catalogue record instead.

I hope this doesn't belabour the point too much - I'm finding it a bit
tricky to get my head around this.

cheers,

Libby


p.s.:

sample rdf: http://yaddle.ilrt.bris.ac.uk/~libby/rec.txt
picture 
http://yaddle.ilrt.bris.ac.uk/~libby/rdfviz-2000-10-23_11-15-54.gif
(from rdfviz
http://www.ilrt.bris.ac.uk/discovery/rdf-dev/rudolf/rdfviz/)




On Sun, 15 Oct 2000, Phil Cross wrote:

> >
> 
> Ralph,
> 
> I think we are in broad agreement, looking at your Annotation schema
> slide (http://www.w3.org/2000/Talks/www9-annotations/slide6-0.html) and
> your paper on the annotation system.
> 
> The obvious difference that strikes me though is that in this diagram,
> all the info, ie the content that comprises the substance of the
> annotation, is contained within a separate document, linked to the
> annotation with the body property. This is clearly useful in creating
> links between separate documents (you could have, as the body, a url
> that pulled out a catalogue record from a subject gateway, I guess).
> 
> However, for our project we're producing a metadata set for evaluating
> health info Web sites, and I think we see these metadata as also being
> properties of the annotation object, in addition to the metadata _about_
> it. (Although after subsequent discussion with Libby Miller, who has been
> developing our RDF comment store technology here, I'm not so sure about this -
> 
> she'll be sending in her own comments on this later).
> 
> It seems then that there are these two approaches - the information
> comprising the annotation being a separate document linked to an
> annotation object using the body property; or the information consisting
> of separate bits of metadata, each being a property of the annotation
> object. These approaches seem to me to be compatible in terms of the
> underlying understanding of what the annotation object actually
> represents, that is in both cases the metadata or the separate document
> are simply constituents parts of the overall annotation.
> 
> Phil
> 
> PS I've been chatting to Libby Miller, who has been developing our RDF comment
> store technology here, about this and she'll be sending in a further reply to
> this .
> 
> >
> > From: Ralph R. Swick (swick@w3.org)
> > Date: Thu, Oct 12 2000
> >
> >    * Previous message: Phil Cross: "Types of annotation"
> >    * In reply to: Phil Cross: "Types of annotation"
> >    * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> >    * Other mail archives: [this mailing list] [other W3C mailing lists]
> >    * Mail actions: [ respond to this message ] [ mail a new topic ]
> >
> >   ------------------------------------------------------------------------
> >
> > Message-Id: <200010121801.OAA16335@tux.w3.org>
> > Date: Thu, 12 Oct 2000 14:00:24 -0400
> > To: Phil Cross <phil.cross@bristol.ac.uk>
> > From: "Ralph R. Swick" <swick@w3.org>
> > Cc: www-annotation@w3.org
> > Subject: Re: Types of annotation
> >
> > Hi, Phil.
> >
> > At 04:28 PM 10/12/2000 +0100, Phil Cross wrote:
> > ...
> > >I feel it is necessary to somehow differentiate between these different
> > >annotation types, so that it is easy to see what is available for a
> > >particular site, and separate annotations by type when presenting them
> > >to users; the question is, what is the best way of doing this?
> >
> > You give a nice concise motivation for the question.  I'll open
> > with the caveat that there might not be one "best".  We can
> > share our ideas and see which of them survive the test of
> > implementation and time.
> >
> > >The basic structure for the RDF triples of comments or annotations, that
> > >we currently use, is to have an annotationID as a unique ID, which has
> > >properties such as 'annotates' (for the URI of the Web page), DC.Title,
> > >DC.subject, etc.
> >
> > That sounds nearly identical to what we've been doing in
> > an annotation service design here in a "live early adoption"
> > project at W3C.  We talked about it at WWW9 DevDay and
> > have continued to work on it as time permits.  We're very
> > close to being comfortable showing it more publicly again.
> >
> > "The W3C Collaborative Web Annotation Project ... or how to have
> > fun while building an RDF infrastructure"
> > http://www.w3.org/2000/Talks/www9-annotations/
> > announced in
> > http://lists.w3.org/Archives/Public/www-annotation/2000JanJun/0023.html
> >
> > >  I'm wondering if we should add an extra property called
> > >annotationType, and try and define some standard values for this. Or
> > >perhaps the 'annotates' property itself should have subclasses (e.g.
> > >'comments' or 'evaluates', or should the annotationID be subclassed?
> >
> > We have defined a generic annotation class and then subclasses
> > of that to declare 'types' of annotations.  We use the rdf:type property
> > to convey the class of a specific instance of an annotation; that is,
> > the semantic intent of the author when the annotation was created.
> >
> > We believe that annotations can in general have multiple types; that is,
> > a single instance of an annotation can simultaneously belong in
> > several classes.
> >
> > -Ralph
> >
> >   ------------------------------------------------------------------------
> >
> >    * Previous message: Phil Cross: "Types of annotation"
> >    * In reply to: Phil Cross: "Types of annotation"
> >    * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> >    * Other mail archives: [this mailing list] [other W3C mailing lists]
> >    * Mail actions: [ respond to this message ] [ mail a new topic ]
> 
> --
> Phil Cross
> Institute for Learning & Research Technology
> Bristol University
> 8-10 Berkeley Square
> Bristol BS8 1HH
> 
> Email: phil.cross@bristol.ac.uk
> Tel.: +44(0)117-928-7113
> 
> 
>

Received on Monday, 23 October 2000 10:16:47 UTC