RE: Class of an annotation body

Hi Jacob –

Just to be clear, I am certainly not suggesting that the vocabulary for the use case I described would be created by this group, or would be part of the OA vocabulary. Not suggesting that we develop an ontology of body types, these will be extensions.

   So for example, there is a “Bird Description” ontology; http://bd.example.com/



<http://bd.example.com/wtp/anno1>     a      oa:Annotation ;

     oa:hasTarget        <http://bd.example.com/wtp/>

     oa:hasBody       [ a bd:Description;

                                                             bd:call  [ a dctypes:Sound; ….] ;

                                                             bd:song  [ a dctypes:Sound; ….] ;

                                                             bd:photo [ a dctypes:Image; ….] ;

etc.



So, yes, as you suggest this is a best practice discussion.  Question is, is the a good practice?



Ray


From: jgjett@gmail.com [mailto:jgjett@gmail.com] On Behalf Of Jacob Jett
Sent: Monday, March 09, 2015 4:58 PM
To: Denenberg, Ray
Cc: Web Annotation; Robert Sanderson
Subject: Re: Class of an annotation body

Hi Ray,

This seems like a best practices type of discussion. You could simply make three different annotations with distinct motivations or make a single annotation with multiple bodies. If I understand correctly, it is this last option that is the crux of the issue.

The question is whether or not we need to develop an ontology of body types for OA to (likely) support various search and selection features. The problem is that it is impossible to think of every possible scenario for typing resources in this manner. The solution in the past has been to leave this kind of functionality to various communities to build extensions to the base OA vocabulary for.

IMO, I think this is the correct way to go. That way we are not in the odd position of mandating certain interpretations of resources (i.e., that an audio clip contains the audio of a bird call or a bird song). Knowing that the resource is an audio clip is just generally useful; it suggests what kinds of software is needed to interact with it. Naming it as a kind of content seems over-specified though and is the kind of information that is likely only going to be interesting within certain communities. Hence it would be best to leave this kind of ontological extension up to specific communities who will be more knowledgeable than us anyway. They can serialize the additional information with the annotation (the beauty of RDF and JSON-LD) and individual consumers can thereafter decide on their own whether or not they should exploit it.

Regards,

Jacob



_____________________________________________________
Jacob Jett
Research Assistant
Center for Informatics Research in Science and Scholarship
The Graduate School of Library and Information Science
University of Illinois at Urbana-Champaign
501 E. Daniel Street, MC-493, Champaign, IL 61820-6211 USA
(217) 244-2164
jjett2@illinois.edu<mailto:jjett2@illinois.edu>

On Mon, Mar 9, 2015 at 3:16 PM, Denenberg, Ray <rden@loc.gov<mailto:rden@loc.gov>> wrote:
Thanks Rob. I think my use case does not fit either of these two.  You tell me if you think it is a legitimate case; I think it is. I’m trying to develop it as a protocol use case but first I’ll describe it in plain English and if it is deemed legitimate, I’ll develop it further.
Let’s say you have an Online Bird Guide -  something along the lines of the Cornell’s service http://www.allaboutbirds.org/guide/search but it relies on the contributions of bird enthusiasts/experts who photograph birds, record their songs and calls, and contribute the images and sound clips.  So, I photograph a White-throated Sparrow in my backyard, in early Spring, and record both its song and its call.  I contribute this information by annotating the service’s resource for White-throated Sparrow: two sound clips(song and call), an image, geographic location (DC area), time of day (morning), period of the year  (early Spring).  And note, these things – location, season, time of day – this is not provenance, this is descriptive information because these songs and calls vary by location, time of day, and season.
I don’t think this fits option 1, because there you have only a single document, a review, and you can signify this via motivation.  In my case, you have multiple documents, for example, two sound clips – one for the song and one for the call – and you have to say which is which, and the only way I can think of to do that is via properties. I don’t think it fits option 2 either, for similar reasoning.
Ray


From: Robert Sanderson [mailto:azaroth42@gmail.com<mailto:azaroth42@gmail.com>]
Sent: Monday, March 09, 2015 3:46 PM
To: Denenberg, Ray
Cc: Web Annotation
Subject: Re: Class of an annotation body


Hi Ray,

The issue is mostly one of provenance and graph boundaries, so falls under the same discussion as named graphs for bodies discussion that we had on a recent call.

There are two main options for body resources with structure:

1.  The body is an external or embedded document containing structured data, or a named graph containing triples, such that the document/graph can be maintained and referred to separately from the annotation.  For example, a review in XML that had fields with number of stars, text of the review and so on fills the same role as a named graph containing the same information with triples instead of XML elements.

2.  The triples are part of the annotation graph, and thus cannot be referred to separately.  Consider as a parallel, the body-as-string versus body-as-resource distinction.  This is the body-as-string case, where there isn't a distinct resource to refer to... there are just triples in the annotation.  For example, the body could be a blank node with numberOfStars and rdf:value predicates.  These assertions are made by the agent that asserts the annotation, which may not be correct ... but also may not be a concern, depending on the use case.  For the general case, and hence for interoperability, the boundary of the body is another challenge.  The set of resources that are needed to make a useful body is arbitrary and use case dependent.  For dedicated systems this may not be an issue, but if you imported the annotation content into another system, it's unlikely that it would know which triples should be included and which not.

Others please do weigh in :)

Rob



On Mon, Mar 9, 2015 at 12:05 PM, Denenberg, Ray <rden@loc.gov<mailto:rden@loc.gov>> wrote:
In the OA data model, there is no general RDF Class defined for the body.  But that doesn’t mean that the body cannot have an RDF class, for certain types of annotations.  For example, for  a tag, the body has class oa:Tag and for a semantic tag, oa:SemanticTag.   (Those are the only two I can think of from the model, maybe there are others I missed.)
There is at least one use case that calls for a structured body and thus an RDF Class (Structured Review:  http://www.w3.org/TR/dpub-annotation-uc/#comment-on-publication-title).   And it also seems that there is a certain amount of hand-waving that suggests that bodies are going to need to be structured in many cases but I don’t think we have had direct discussion of this.
I am in the process of developing a use case along these lines and would like opinions on whether this approach is seen to be legitimate or frowned upon – annotations where the body has properties - because there was discussion of this in BIBFRAME and it seemed to cause controversy.
Ray





--
Rob Sanderson
Information Standards Advocate
Digital Library Systems and Services
Stanford, CA 94305