| Patrick et al,
Occurs to me others form ODF meta might like to see these files as well.
This is the presentation and model document that I presented at the Anatomic Pathology Informatics conference (APIII) in Pittsburgh a few weeks back.
My presentation is here. I did it on Mac, so those who want to see the original Keynote version can download it at
Powerpoint, pdf and Quicktime versions of the presentation, together with the sample document that shows how an XForms in an ODF document can be rendered in RDF and potentially submitted (see comments to Patrick below) are at the other link:
There was a lot of enthusiasm for this there, but a lot of folks there are non-ODF'ers so I think some of it went a bit past them in a 10-minute presentation. Nevertheless, I got many positive comments.
Cheers!
John
On Oct 12, 2007, at 12:16 PM, John F. Madden, MD, PhD wrote: Hi Patrick,
The materials from my presentation are downloadable at:
It's a big zip file because I included the presentation as ppt, pdf and also as a quicktime movie because the animations in the Keynote version didn't translate well to either ppt or pdf.
The ODF document is a simulated pathology report that generates an RDF file from the filled-out XForms in the document. It does it through an unnecessarily complicated mechanism, as I wrote this document so as to support both a conventional "concrete" XML form data representation as well as the RDF form data representation, and the two alternative representations are "intertwined".
When I redo this document -- as I will sometime soon -- I will remove the concrete XML and put all the data directly into RDF.
To add this feature to OpenOffice, you would merely have to support a form submission action that would save the generated RDF to the ODF document package itself (rather than posting it to a url or saving it as a local file, which are the currently supported submit actions). I don't think there is a way for a user to create such a submit action in OpenOffice, but it should be a totally trivial job for one of the Sun programmers. The barrier is just that the ODF document is zipped, and it is hard to write directly to a zipped file !!!
John
On Oct 12, 2007, at 8:37 AM, Patrick Durusau wrote:
John,
Thanks for the explanation!
That makes sense. It isn't going to cost implementers to have the additional baggage and as you say, it may make it more attractive to the machine reasoning folks.
I don't see any problem with it appearing in the final version that makes it into this version.
BTW, working on the opening keynote for the conference in the Netherlands and have settled on using the work of Grotius (a native of the Netherlands) on open seas to make the same arguments for open document formats, enabling commerce, etc. Should be interesting to see how it comes out in the end.
A natural law argument that if users own their documents (who else?) then they necessarily should have the freedom to choose to process them with any application they prefer. Following on from that, enabling that freedom of users means that there will be more commerce and competition in applications to process those documents.
Hope you are looking forward to a great weekend!
Patrick
John F. Madden, MD, PhD wrote: Patrick,
It's an entirely reasonable question to ask why you would ever want to reason over them, and maybe you wouldn't.
But the Semantic Web principle is that you never know how somebody might want to reuse the information you post in unanticipated ways. And if you define your concepts well, then your ontology is going to be "better liked" by people, more apt to be reused, and people are hence more likely to adopt your ontology.
There is a trade-off, of course. If you put *too* much detail in, then conversely people might dislike your concepts because they are *too* tight, *too* context-dependent, *too* tied to their origins and for this reason unfriendly for re-use.
So that, I think, is the art of the Semantic Web. I think if we put in a few modest, uncontroversial disjoints, and if we state the obvious domain/range constraints, then that, in my opinion, is a net plus.
Incidentally, in my experience there are two kinds of semantic webbers. There is one group that wants to use semantic web just to support searching; there is another group that wants to use it to support machine reasoning.
The search people basically just want concept linkages connected up in a big network, so that their search engines can walk through the link networks and harvest potentially relevant facts and documents along the way. These people do not care much about disjoints, domains and ranges. They tend to be happy with pure RDF, and they don't find any advantage to using OWL.
But pure RDF has such a limited set of built-in logical constraints that pure RDF reasoners are extremely puny, capable of very little useful inferencing. So the machine reasoning people, who don't care so much about searching, but instead want to extract inferences from a pre-harvested set of facts, prefer OWL over straight RDF. And accordingly they like to be much more explicit and detailed in their class and property definitions, in ways which are only possible under OWL.
Like I say, the art of the thing is to try and satisfy both constituencies.
John
On Oct 11, 2007, at 4:57 PM, Patrick Durusau wrote:
John,
John F. Madden, MD, PhD wrote: Patrick --
Exciting news about your upcoming talk. I'm going to send you separately my wonderful (if I do say so myself) example from my September conference, which is an ODF doc that uses XForms, and generates an RDF/OWL representation of the form content to be reposited as ODF metadata. I'll also send you my slides that you are welcome to use. Do you happen to use a Mac (my presentation works best in Mac Keynote, but I can make it into a PowerPoint if that's what you use).
Great!
Powerpoint as I am not on a Mac. w.r.t. the other issues:
While cleaner from an ontology design perspective, what ability do we gain from the disjointedness axioms? While I don't disagree that they are pretty obvious, I am not sure what adding the axioms would mean in terms of actual processing. There may well be some real advantage that I am not seeing so please understand this is truly a question and not opposition to your suggestion.
The advantage is that if anyone ever wants to perform reasoning on the information using an OWL reasoner, then the reasoner has more information to work with and can therefore make more (and more accurate) inferences than would otherwise be possible.
Remember that an OWL reasoner makes NO prior assumptions WHATSOEVER about disjointness of classes. It's a common source of problems for neophyte users. They'll create some elaborate ontology of, say, cats. They'll have subclasses Lions, Tigers, Housecats, etc. But they won't explicitly state that these classes are disjoint. Then they try to reason on the ontology and they get all kind of crazy inferences, like that some Housecats live in the jungle and eat villagers, and some adult tigers weigh under 3 pounds and like to eat FancyFeast out of a can.
Even worse, if they have a class Dogs and they don't explicit;y declare it disjoint from Cats. Or ---horrible to imagine -- Cars, Stars, FamousPhilosophers, RockFormations, GastrointestinalDisorders, Cats and Dogs. Mama mia!
I've forgotten what I suggested should be declared disjoint in the ODF ontology. Remind me.
Nothing all that remarkable:
Package, File and Element are all mutually disjoint. ContentFile, MetadataFile and StylesFile are all mutually disjoint.
What was puzzling me was why anyone would use a reasoner over them? Granted you could, but it looks like overkill to me.
Still, it doesn't cost us anything so perhaps we should.
On the notion of domains and ranges, is this a question of specifying what qualifies as legitimate values? That is you want to declare what are acceptable values? For validation purposes? Again, just not sure what this adds so asking for more information.
This is similar to above. If you don't declare domains and ranges, then an OWL reasoner will assume that the subject and object of the value can be members of any class.
Actually, in OWL, the "meaning" of a property -- for purposes of automated reasoning -- is entirely defined by its domain and range restrictions (and by its subPropertyOf relationships).
So say you have a property hasHairColor in your ontology, but you don't declare domains and ranges explicitly. Then an OWL reasoner might come up with a "true" inference, given appropriate starting facts, that John'sTelephone hasHairColor Patrick'sEyeglasses.
Obviously, the reasoner doesn't "understand" the "meaning" of the property. You have to have told it explicitly that the hasHairColor property has a domain of things in the class People, and a range of things in the class PossibleHairColors. Then it will only generate inferences that are valid for the property used as intended.
Ah, ok, now I understand.
Thanks!
Hope you are having a great day!
Patrick
John
-- Patrick Durusau Chair, V1 - US TAG to JTC 1/SC 34 Acting Convener, JTC 1/SC 34/WG 3 (Topic Maps) Co-Editor, ISO/IEC 13250-1, 13250-5 (Topic Maps) Co-Editor, OpenDocument Format (OASIS, ISO/IEC 26300)
-- Patrick Durusau Chair, V1 - US TAG to JTC 1/SC 34 Acting Convener, JTC 1/SC 34/WG 3 (Topic Maps) Co-Editor, ISO/IEC 13250-1, 13250-5 (Topic Maps) Co-Editor, OpenDocument Format (OASIS, ISO/IEC 26300)
|