Re: Associating context with an annotation

Hi James,
( as suggested by Robert, I will break things down to the specific 
issues and topics on this list to annoy you on a lower level ;) . I even 
promise to use OA jargon as often as possible. )

I am unable to understand the purpose of context in Option 2. I think it 
should be clarified, why you would need a context  for an aggregation of 
annotations as this is not mentioned in the scenario on top of the page.

As a comparison: the context model in NIF serves mainly these purposes:
1. limit the things that you can say about a selection (words are highly 
ambigue, depending on the granularity "house" can have dozens of 
meanings http://en.wiktionary.org/wiki/house#English-abode  ). Thus 
"house" in a certain context is less ambigue and can not have all 
readings any more.
2. model context change, i.e. which annotations stay valid, when the 
context changes.
3. allow to automatically reason, whether two selectors select the same 
thing. This means "physically" inferring owl:sameAs between the 
respective Selectors and SpecificResources. (immensely useful to answer 
queries)

For 1. If no explicit reference is given, NIF always implicitly assumes 
that the context is the whole document. This is consistent to an 
annotator, who marks a text and adds a DBpedia URI to a specific 
selection. The annotator could annotate the text "Barack" because he 
knows the document is talking about Obama.  The current restriction is 
that the selection and the context have the same format. This allows to 
encode an "in" or "part of" meaning into the "hasContext" property, 
which is really helpful for machines. Thinking about annoting a video in 
a blogpost. It seems strange to me to only annotate part of a video in 
part of the document.  So when having different media, I would assume 
that we only want to annotate "whole" things such as a "whole" video in 
a "whole" blogpost.  Note that I don't mean that you can connect a part 
of a text to part of an image. I am just saying that it might not be a 
hasContext relation.


2. The second part might not be relevant for OA.

3. The third part is done with two owl:hasKey axioms. Are there any 
formalizations available for OA ? Did you define, when two 
selectors/target are the same? I think the OntoClean approach might help 
here (somebody wrote a very acceptable Wikipedia Article 
http://en.wikipedia.org/wiki/OntoClean). Personally, I am quite 
conservative and I favor extending expressiveness only, if it is still 
given, that you know exactly how you can conclude whether two 
annotations refer to the same target.

All the best,
Sebastian


Am 17.08.2012 15:57, schrieb James Smith:
> I've added a page (http://www.w3.org/community/openannotation/wiki/Annotating_Resource-in-Context_Proposals) linked off of the Open Issues page. Not sure it's in the right place or has the right title, so feel free to edit. Feel free to add to it.
>
> The page briefly outlines the problem, a scenario, and two ways in which we might associate a context with an annotation.
>
> -- Jim


-- 
Dipl. Inf. Sebastian Hellmann
Department of Computer Science, University of Leipzig
Events:
   * http://sabre2012.infai.org/mlode (Leipzig, Sept. 23-24-25, 2012)
   * http://wole2012.eurecom.fr (*Deadline: July 31st 2012*)
Projects: http://nlp2rdf.org , http://dbpedia.org
Homepage: http://bis.informatik.uni-leipzig.de/SebastianHellmann
Research Group: http://aksw.org

Received on Friday, 17 August 2012 15:39:24 UTC