Thanks Kevin for this attempt - I, at least, found it helpful in terms of thinking about how BIBFRAME uses 'annotations'.
There is clearly a lot of 'grey area' in terms of what might be regarded as an annotation and what not. While acknowledging the criticism others have made that dealing in 'facts' is problematic, I think I understand the idea that there are things that we might easily get consensus on and those that we might not, or have valid multiple views on - but this is a spectrum.
In this context it makes sense that Annotations are used when it is important to the provenance (especially the 'who') of a statement. The issue of Provenance is mentioned by Rob Sanderson as part of the rationale for developing additional vocabulary/ontology for the Open Annotation work in one of his recent emails. The other work on Provenance for triples has also been mentioned, but clearly the current situation is that OA creates a clear mechanism for this)
The other examples of requirements from the Open Annotation work given by Rob were (numbering assigned by me here):
1) A highlighted span of text. There is an obvious target segment of a resource (the object), but there is no body/comment (the subject). As a triple must have a subject, this could not be expressed. A second example of this would be a bookmark where the body is also implicit.
2) An annotation that refers to multiple segments of a resource, multiple resources or multiple segments of multiple resources. In this case there would be multiple objects, which is also not possible to be expressed in RDF.
3) Where there are, equivalently, multiple comments, such as a comment in English and the same comment in French and the user agent should determine which is more appropriate to show to the user.
While not an expert, (1) and (2) seem clear to me. I'm less clear why (3) can't be handled as a language tag - although I would see an argument that a translation is in itself an annotation of a kind :)
I don't see anything in Bibframe equivalent to (1) - all bibframe annotations are intended to point at a resource, not a fragment as far as I can see?
I don't see anything in Bibframe equivalent to (2) - all bibframe annotations are intended to point at a single resource, not multiple resources as far as I can see?
I think (3) could apply to Bibframe - specifically in terms of bf:coverArt
Having thought this through for me the first question is whether all Bibframe 'annotations' as currently proposed should be expressed as annotations. I think I'm willing to accept that if they fulfil any of the criteria above (provenance is key, or 1,2,3 listed) then there is a justification to use an 'annotation' approach.
bf:Review (provenance is key)
bf:CoverArt (assuming this is an image of the cover - equivalence between different images of the cover but all ultimately making the same assertion of 'it looks like this')
The others seem less clear to me:
bf:ContributorBio - maybe if provenance is key, although it may depend on the type of biographical information being asserted - feels like it needs breaking down further
bf:TableofContents - is this debated? Or likely to need multiple equivalent assertions? Feels like this is just a straightforward property of the work
bf:SampleText - while I can see that there could be many examples of 'sampletext' for a single item, it doesn't seem likely we care 'who' made the claim it was sample text? Feels like a different kind of relationship to an annotation
bf:PublisherDescription - this feels wrong in that why not have bf:Description, with the annotation asserting it was created by the 'publisher'? A 'description' seems squarely in 'annotation' territory, while a specific description assigned to a specific body feels like it could be handled without resorting to annotation
Anyway - I guess my first question (I have a second for a separate email!) is - for each case where annotation is being used at the moment in BIBFRAME, does it really make sense, and if so, why?
Owen Stephens Consulting
Dear Karen, all,
In reading your email (the below and others) as well as one or two emails from other individuals, it became clear that we missed the forest for the trees when it comes to basic definition. So, I wanted to offer up an answer to the basic question "What is a BIBFRAME Annotation?"
I make no claims to have addressed all of the questions you raise, but I wanted to start with the basics before moving on to more specific details, such as whether BIBFRAME Annotations are end-user-oriented or cataloger-oriented, which is a question you asked in a separate email I believe. Naturally, if this spawns additional questions, please ask.
What is a BIBFRAME Annotation?
A BIBFRAME Annotation is a resource that enhances our knowledge about the resource it annotates (the target resource). A BIBFRAME Annotation manages this in one of two ways. One way is for the BIBFRAME Annotation to facilitate an association between two resources by means of relationships. The resource being annotated is the target of the annotation while the resource that otherwise enhances our knowledge of the target resource is the body, or payload, of the annotation. The BIBFRAME Annotation, in this case, serves to say, "Resource A annotates Resource B, the target resource." The other way is for the BIBFRAME Annotation to be itself the carrier of additional information about the target resource. In this alternative, the BIBFRAME Annotation does not function as a lightweight abstraction layer bridging two resources, but an end resource that further enhances our knowledge about the target resource. As a matter of focus, a BIBFRAME Annotation generally refines our understanding of the target resource as a whole versus any one particular aspect or segment of the target resource.
Another distinguishing characteristic about a BIBFRAME Annotation, as distinct from a BIBFRAME Work or BIBFRAME Instance, is that *who* asserted the Annotation is of paramount importance. The *who* being the agent stating, "This annotates that." The importance may range from simply wanting to know, for the sake of completeness, the source of the added information to needing to make a value judgment predicated on the identity of that source. The latter is particularly meaningful when the additional information may be subjective in nature. Reviews and ratings fall squarely into the realm of subjectivity, where knowing *who* is asserting the value of the review (not to mention the identity of the reviewer) may directly inform how the Annotation is treated.
Another way to define a BIBFRAME Annotation in this regard is by contrast to other BIBFRAME resources. In the BIBFRAME universe, most of the "facts" about BIBFRAME Works, Instances, and Authorities are immutable, and they will likely be of interest to most users. Creators, producers, authors, editors, places of publication, publication dates, publishers, manufacturers, titles, and much more, do not change per individual resource. The novel /The Heart of Midlothian/ by Sir Walter Scott will always be titled "The Heart of Midlothian" and be by Sir Walter Scott. Likewise, the instance published in 1878 in New York by G. Munro cannot shake those facts in just the same way the instance published in 1885 by J.W. Lovell and company (also in New York) cannot escape from those facts. These are unquestionably objective "facts" about those resources. However, reviews of the Work will be subjective and come from a myriad of sources, some of which may be more trusted than others or may be more suitable to some audiences than others.
Using the BIBFRAME Annotation model for game ratings provides another example. The Entertainment Software Rating Board (ESRB) assigns ratings to games based on content and age. Australian Classification Board does this for the Australian market. The Computer Entertainment Rating Organization (CERO) is the Japanese equivalent of ESRB. The South Koreans, Germans, Europeans, and many more groups have their own rating systems. In the United States, Common Sense Media is an alternative rating system that places special emphasis on age appropriateness. Rating systems abound and, despite similarities, each will be distinctive to their markets and audiences. Not only do the systems, by their nature, proffer subjective evaluations of media content, their value is only fully realized when we know *who* has assigned a particular game rating. The BIBFRAME Annotation model provides a flexible way to enhance the description of Works and Instances while enabling a scenario that maintains a certain separation between objective "facts" and subjective ones.
The valuable information added by the BIBFRAME Annotation is, objectively, no less (or more) important than the information associated directly with a BIBFRAME Work, Instance, or Authority. It is just that the information conveyed by means of a BIBFRAME Annotation is in enriched by knowing *who* asserted the BIBFRAME Annotation. As a practical matter, the *who* in this model can become a filter, allowing consumers (libraries certainly, but potentially also patrons) to select annotations based on who asserted the annotation.
I should add that we believe the BIBFRAME Annotation model to be a positive development that will allow for a fair amount of flexibility in the future for libraries, and other implementers, to augment their data how they deem most appropriate while leaving the information that remains constant between descriptions untouched.
We still continue to explore the possibilities and potential of the BIBFRAME Annotations within the BIBFRAME model as a whole, so we appreciate the additional eyes and questions - it is about identifying and enabling our use cases.
From: Bibliographic Framework Transition Initiative Forum
[mailto:[email protected]LISTSERV.LOC.GOV] On Behalf Of Karen Coyle
Sent: Friday, May 03, 2013 3:35 PM
To: [log in to unmask]
Subject: [BIBFRAME] BIBFRAME annotation
I've had a terrible time trying to understand Open Annotation (why is
this not just an RDF graph showing a relationship between things? Why
does it get its own formal definition?), and now I'm looking at
BIBFRAME annotation, pretty much guaranteeing even greater confusion on
BIBFRAME annotation is described as:
The parties and objects involved in a BIBFRAME Annotation are:
. The Target of the Annotation: A BIBFRAME Work, Instance, or Authority.
The book, in part 1 of the illustration below.
. The Annotation Body, which is the payload of the Annotation. The book
. An author, artist, reviewer, etc. who writes the Annotation Body.
(This role is not represented formally in the Annotation model, but is
mentioned here to clearly distinguish it from the Annotator.) The
. The Annotator, who asserts the Annotation. (The Annotator is not
necessarily the same party as the author, etc. who wrote the
Annotation.) The Annotator in part 2 of the illustration.
. The Annotation itself , which points to the Body, Target, and
Annotator. The Annotation, in part 2 of the illustration. [1 - section
From this description I conclude that "Annotation" is a special
instance of "node" -- a node with some semantics and a limited set of
properties: links to a particular set of things. I'm still totally
unclear why this is a special case in RDF, since things and links to
things are inherent in the model.
What BIBFRAME seems to be doing is using Annotation to mean "optional
information." I conclude this from section 2.1 of the BIBFRAME
annotation document [1 - section 2.1]:
What is a BIBFRAME Annotation?
For purposes of this model, a BIBFRAME Work, Instance, or Authority is
an abstract resource. Different institutions may have different views
of any given BIBFRAME Work, Instance, or Authority. For example, for a
given BIBFRAME Work, InstitutionA and InstitutionB may each have a view
of the Work, bf:Work A and bf:Work B.
Certain information is integral to a Work - title and author, for
example - and might be reasonably expected to be reflected in both
views. Other information might be part of one view but not the other -
information asserted (possibly by a third party) about the Work, which
Institution A chooses to integrate into its view but Institution B
chooses not to (or vice versa).
A BIBFRAME Annotation is an assertion, by any party, about a BIBFRAME
resource (Work, Instance, or Authority) that any institution holding a
view of that resource may choose to integrate into its view, or choose
There seem to be two things going on here. One is that different users
of BIBFRAME will make different choices about what is "integral" to
Work, Instance and Authority.
The other thing is that there are *optional* bits of information that
can be encoded as Annotations, and these can be ignored by anyone not
interested in making use of them. Unfortunately, defining some elements
as "unessential" means that others must be defined as "essential."
This means that one person's "integral bit" with be another person's
Annotation. Thus having annotations doesn't mean simply that you can
ignore all Annotations, nor does it mean that you do not need to make
choices among the "integral bits" that come from other sources. In this
sense, Annotation doesn't appear to me to solve the problem of
differences in cataloging.
I *could* understand (although not necessarily favor) a regime in which
there is a defined core (oh, yes, that word again) and everything else
is an annotation. That is, everything else is optional. But the
definition of Annotation here does not seem to make this separation.
Another possibility for Annotation would be to define it as being
"third-party information" -- anything not provided by the cataloger and
not provided for in the cataloging rules. I'm not saying this would be
a good idea, but it would be a clear separation between Annotation and
If there isn't some clear separation, then I don't see a great
advantage over letting metadata users select elements based on data
elements and provenance.
What have I missed?
[log in to unmask] http://kcoyle.net