[162]                             home                             [164]

 

Friday, February 10, 2006

 

Challenge problem à

 

Generative Methodology Glass Bead Games

 

On using RDF to model web services

 

 

 

From: protege-owl-bounce@crg-gw.Stanford.EDU

[mailto:protege-owl-bounce@crg-gw.Stanford.EDU]On Behalf Of Rinke

Hoekstra

Sent: Friday, February 10, 2006 8:35 AM

To: protege-owl@smi.stanford.edu

Subject: [protege-owl] Re: importing RDF

 

Footnotes are short responses to specifics

Longer reply is at à [164]

 

 

Dear Paul,

 

I really hope this mail clarifies [1] some of the things that have been bothering you over the past few days. Hopefully thereby quenching that which is (in my humble opinion) gradually turning into a flame war against the innocent (on both sides!).

 

Paul S Prueitt wrote:

> I am absolutely stopped on something that I need to finish ... this  (really really should be simple) importing of an Altova created simple  RDF file.   [2]

 

Please, see my other response on your original problem report. (I hope it suffices) [3]

 

>  Everyone knows that I am critical of how very complicated the ontology > paradigm is that has been created by the Stanford team, and I hope that I do not get attacked (again) by someone who feels that I am being unfair to anyone.   This attacking response in face of criticism is simple NOT proper.

 

First off, I believe your fairly aggressive criticism of Protege is often indeed unfair. Some people have a natural reaction to react in the same tone if they believe something they have a high regard for is criticized in this way. But, yes... let's keep the discussion pleasant ;)

 

The KR paradigm used for Protege hinges on two representation languages: Protege-Frames (developed by Stanford, and indeed 17 years old), and the OWL language (developed by esteemed scientists, and a W3C recommendation, only several years old). This distinction has often been a source for confusion (especially in loading files, such as your BioPax problem), of which the Protege team is aware. Unfortunately some bugs/features have a higher priority than others.

 

The ontology paradigm that you are talking about (I assume it's OWL), is based on description logics. The people at Stanford and Manchester took it as their mission to build a tool for this complex language, that would make ontology development relatively easy (compared to hand-coding). As starting point, they took their well established Frames editor, and turned it into the OWL powerhouse it is now. However, OWL still is no easy language...

 

> FIRST the import of a compliant short RDF file into Protege, or the import of the BioPAX file into Protege should not be a problem.  Period.  But there have been at least four different persons who have had a problem importing the BioPAX file.  This is not the BioPAX working group's fault.   It is the fault of the Protege GUI.  Right?

 

Yes, to some extent it is: the distinction between Protege Frames and OWL is not clear to everybody. Please keep in mind that the OWL support for Protege is developed as a plugin. The core protege is still frames-based (CLIPS).

 

> The design of the import interface seems to be fundamentally flawed.    Why, after 17 years?

 

It really is fairly young (younger than OWL is), actually. And I agree the process of importing OWL files could be more clear. I believe Matthew Horridge is currently developing support for double-click opening of owl files.

 

> The flaws might stem from the Tim Berners-Lee Layer cake itself, and so  the difficulty of the import and export may not be merely a result of  poor work by the Stanford team.   But the problems themselves are not  being exposed, (I conjecture).  

 

The 'flaws' are the result of many things really, but the layer cake is not one of them. One reason is (again) frames vs. owl, and the other is the apparent refusal of many owl/protege users to RTFM. OWL is a complex language, yes, and it requires more in-depth knowledge of knowledge representation and its formalisms than say, UML. (and note that XMI is non-standard)

 

>  > In my message, two days ago, to Protege Discussion (which got rejected automatically because I used the word "OWL") I trying to set up a problem, modeling the OASIS SOA IM with RDF !!!or!!! OWL.   This task is huge, because it links the OASIS and IEEE web services and XML standards with semantic web "ontology with inferencing".  So, I am asking for help.

 

Renaming mailing-lists is not always a good idea, but protege-owl should really be called 'the ultimate mailing list for those who want to do something roughly semantic-webby with Protege (including rdf/rdfs and ofcourse owl)'. The protege-discussion mailing list is really only for frames users.

 

> The RDF modeling of OASISA SOA IM task should be a great interest to everyone on this discussion forum, as well as the Protege-discussion forum (where it never got posted).    Can I see a show of hands?  Who in the Protege OWL forum would like to see a simple RDF model (no OIL) of > the SOA IM?  Once we have this, who would like to see various ways of adding the OIL?

 

I abstain ;)

 

> I want to be clear, since over and over my discussion is treated as if I  need for tutorial on something (name edited)

 

Perhaps this is not such a bad/insulting idea. Please bear in mind that *any* tool has its specific problems, tweaks and quirks. Reading the tutorial might help to prevent some of the problems you have experienced. This is not because we don't regard you as a 'leading researcher in this field' (your own words: I've never heard of you before personally...), but because we *know* protege is not really mature yet and needs some introduction.

 

> The problem is not my knowledge of the problem.  The problem is on the incompleteness of our (the world society) development of tools for modeling "reality".  This problem, as Andrea and I have talked about starts with the narrow (and often misleading) use of terms like inference, and the absence of common understanding of the generalization of the term "inference" ...  "entailment" is both logical "inference" and physical cause.

 

It is indeed that very problem. But it feels unfair to blame those who are trying to do something about that... even if it is on US *and* EU/UK research money. I am a Dutch citizen ;)

 

> And, as has been suggested by several PhDs in knowledge engineering,  this is not about one professions right to have their own private language.  The entire world needs to have ontological modeling figured  out ... in precisely the way that arithmetic was figured out and is now commonly and everywhere available without "professional interests"  trumping "social interests".   Again, I do not deserve to be attacked as being somehow un-professional.

 

Agreed. This understanding of arithmetic is done through education though: both of the problem and of the formalism (formulas etc.). Wouldn't you agree that this holds for ontology engineering too? Even if the words used may be confusing to the 'uneducated', they might very well be adequate within context, oops here we go again... ;)

 

> It is time to see the limitations and problems of Protege "with eyes > open".

 

Have a look at what's being done at Manchester in the co-ode project. They have taken their experiences in teaching owl/ontology engineering and turned them into invaluable tools and plugins for Protege. The problem *is* taken seriously. Perhaps you haven't really had a chance to  notice.

 

> We all have many layers of things we are trying to get to .... and the fundamentals of import and export using Protege cannot be so weird as to bend one's mind into contortions. Right?   (gesturing with a puzzled look).

 

... please give it some time. [4]It's a two-way street: most of the import problems stem from other tools not exporting the right format. Or people trying to import the wrong kind of file using the wrong wizard.

 

> As I mentioned, the problem in loading the Altova created RDF file is  likely a namespace problem....

 

Indeed it was, plus a problem with Altova producing syntactically correct, but semantically incorrect rdf:about attributes. (semantically incorrect as in not corresponding with the meaning you intended to give them)

 

> but how this is handled should be made clear in the "software wizard" that is now just poorly designed. Right?

 

If the files presented to the import wizard are syntactically correct, and correct with respect to the RDF specification, the wizard has no way of knowing something went wrong. Really. This is just the way computers work.

 

All the best,

 

          Rinke

 

 

 

 

 

 

 

 

 

 

 



[1] This email clarifies a point of view, but does not in any way help address the practical issue of how to create a OWL representation of the OASIS SOA IM standard.  Nor is the email intended to address this issue.

[2] The confusing  behavior of the import has still not been addressed by anyone.

[3] The issue of the un-explained errors in importing a valid RDF file (developed using Altova) was never addressed. The only substantive response is to say that the team will look into the matter. 

[4] (see – [164] )