Re: Mechanisms for expressing "aboutness" effectively?
"Andrew S. Townley" <[email protected]>
| Newsgroups | gmane.text.xml.xtm.general |
|---|---|
| Message-ID | <[email protected]> |
Hi Patrick, Thanks for your reply, and sorry my point was so occluded. Thanks for making the effort to understand what I meant! On 12 Apr 2011, at 11:05 a.m., Patrick Durusau <patrick-Q/[email protected]> wrote: > Andrew, > > On 4/11/2011 7:24 AM, Andrew S. Townley wrote: >> Hi folks, >> >> A modeling question for you which, after much hair-pulling, googling, RDF/OWL ontology analysis and reflection, I still can't resolve to my own satisfaction: >> >> What mechanisms have you all found most useful in expressing the concept of "being about" between two things? >> >> A common approach is to provide an instance-specific view, e.g. applying dc:subject, sioc:topic or similar properties to particular concept instances (where those concepts are article, book, blog post, billboard, song, etc.). However, this doesn't quite give me the modeling flexibility I need as I want to support both declarative and annotative uses with the same construct and without editing/modifying the originals. >> > Ah, I had to read your post several times to see what I think is the issue. > > You mean that dc:subject is too limited to capture things like time of when it was applied, who applied it, in what sense is "subject" being used, etc. > > Yes? > > I am not sure what you mean by "support both declarative and annotative uses with the same construct and without editing/modifying the originals."? Basically, yes. Remember in my environment, merging of proxies doesn't happen by default, so it is important to me where things go in a logical graph structure for these sorts of problems. I generally want the information structure to be as clear as possible about where particular information assertions are created and by whom. The direction and "ownership" of the relationship also is a factor in the way I think about information modelling. With property-based assertions tied to a proxy, the relationship's only really in one direction, and it really represents a "collapsed" view of a more expressive relationship, optimised for the perspective of the person manipulating or viewing the source proxy. As you mention, the expressiveness I'm after isn't available through this collapsed viewpoint. > 1) I don't see a difference between declarative and annotative, unless that is a distinction you want to make for some reason. Say the difference between an author who "declares" some fact versus some later contributor who supplies the same "fact" by way of annotation. > > Quite possibly handling that by attributing "facts" to their originators so declaration/annotation depends on its originator. Part of this relates to the comments above, and part relates to the temporal nature of the way annotations or other information is associated with a particular subject. However, I do want to keep this distinction clear, and I also want to do it in a way that doesn't change or otherwise require write access to the proxies being manipulated to establish these relationships, be they subject bindings or any other typess. Basically, a linking proxy modelled as an association is the normal way I enable what you describe regarding traceability. > 2) By same construct do you mean a common association for both coarse as well as finer grained properties about a subject? > > Say a recipe type association that allows you to say a dish is seafood and then to later say it contains crawfish and even later to say that it is crawfish etouffee?* Well, maybe not for all time, but that's the basic idea. Another principle I've been using for a while is related to iterative disclosure, so that you establish general relationships initially, then either update or augment these relationships with more specialised ones--possibly depending on a shared domain, but possibly also dependent on both your own or others' growing familiarity and understanding of the domain and its relationships over time. Your example is a little different than my original "aboutness" issue, but it's illustrative of the kind of thing I mean for sure. :) > 3) I am not sure what scenario would involve editing a prior association so I am not sure how to answer that concern. Why would you do that? > > Hope you are having a great day! > > Patrick > > *If you are writing a topic map that includes crawfish etouffee, you need to note my wife fixes the best crawfish etoufee to be found anywhere. Just an FYI. ;-) Wow! That's useful information to have. I love Creole and Cajun food. Just made a big batch of Jambalaya on Sunday. Some day we need to organise beers and some spicy food! :) > Thanks again for your comments. ast