Generative Methodology thread

Friday, February 10, 2006

On using RDF to model web services

In off line discussions with Timothy Redmond we have talked about the frame paradigm.

I will summarize here and ask for comment and continuing discussion. I am sure that I have some of my understanding of the literatures incomplete. Corrections are appreciated. I am working hard to change my understanding when this is justified.

Timothy commented about KIF..

<quote>

Kif is aimed at a very high level of expressiveness

(meaning less amenable to reasoning/automated processing). Hence the

slide:

http://www.ksl.stanford.edu/people/fikes/cs222/1998/KIF/sld002.htm

Protege Frames (following the okbc standard) has the goal of being a

more manageable language aimed at ontologies. Kif has very different

goals than ontology languages such as OKBC, Frame Logic, RDFS, OWL,

OBO, etc.

Clearly he is right (and also wrong). He is right because a majority of knowledge engineers think about this in the same way as he does.

The notion of a frame, as expressed by Schank, was not necessarily connected to an "expressive" logic. In linguistics the notion of a frame is similarTo the notion of cognitive context. But in all cases, we have the goal of creating descriptions of reality, eg "ontology". This is the similarity. The difference is in modeling philosophy. Here one needs several approaches, one being the logic based ontology. What are the other approaches?

Here is where one gets into "other schools" about formalism... and the process in which formalism is created (ie as part of human cognitive acts and communication).

I am reading now the works of Alan Rector, regarding what might be regarded as some specific limitations of OWL.

I will leave others to the details..

The paper I reference is: Defaults, context and knowledge: alternatives for OWL-indexed knowledge bases (2003) at

http://www.cs.man.ac.uk/mig/people/rector/alr-papers.html

I am also looking at the work suggested more recently (Sept 2005) by Frank Olken, Kevin Keck and John McCarthy (all at Lawrence Berkeley National Laboratory)

<quote>

Metadata registries are intended to span between conceptual models (semantic specifications, data element concepts) and information artifact specifications (data element specification).

(end quote>

on improving relationship modeling (using ISO/IEC 11179)

Timothy wrote to me:

"The development of new languages and vocabularies is somewhat out of

the scope of the Protege effort. In general, we have usually

followed the standards or w3c recommendations (frames comes from okbc

standard, owl is from w3c and now we are looking at owl 1.1). We are

watching your work and if we see new languages develop that people

find useful, they very well could find their way into Protege."

The point is that a concept of a frame is implicit in the construction of an n-ary, but that "logic" is not implicit in the construction of an n-ary. The n-ary, as I have advocated can be merely an evocative symbol.

http://www.ontologystream.com/aSLIP/index1.htm

Logic, and the use of (to quote Rector) "logic based ontology as an index to contingent information about " ... ) is vitally important to "what is next". (My opinion.)

But the research imposition of incomplete and "results in "descriptive and expressive logics" may cause more confusion that order.

" Beginning with pioneering efforts such as FRL and KL-ONE, interest turned increasingly to logic based representations using the notion of "definition" rather than "prototype". Unfortunately, logic based mechanisms for capturing the notions of default reasoning (non-monotonic reasoning) proved problematic, and all suggested solutions were computational intractable. ...

The viable alternative seems to involve the human domain expert, which means NOT the IT consultant, in the assignment of meaning to "frames".

Again, the problem we have is the terminology used in the knowledge engineering research world is in conflict with how the terminology is used in normal social discourse. So when IT systems are sold, there is confusion in the procurement process and there is no knowledge management terminological reconciliation. The confusion then may result in failures that are very costly in terms of economic and social capital. One community, the community of procurement specialists, does not even know that there is an issue in a divergence of semantics about what "semantics" means.

The point is made, by several knowledge engineers, that the overwriting of terms like "frames" is a normal specialization that any academic discipline deserves to make.

But the larger issue is the one that the group at the National Lab is addressing, and this is to make the mapping between OWL constructions and controlled (or managed vocabularies)... as well as to XML metadata registries.

(I hope I am not misrepresenting anyone's work. Please correct me if I am.)

This is where my work is addressing a specific "stratified" approach where invariance observed is convolved into categories as a process of observation (ie in Rector's terms, "using prototypes that are observed as use cases").

The knowledge engineer’s definition of an axiom, as I have learned from John Sowa and others, is assertive in nature. This is quite different from the mathematician’s definition of an axiom.

For the knowledge engineer an axiom is often an assertion of truth that a class has a specific subclass and that a transitive property is "ontological". In Protégé this is done with mouse clicks. But the point is that most professionals in the knowledge representation field feel that expertise in exotic formal language allows them to make these assertions as part of creating the Semantic Web system that then is imposed on everyone else. I am exaggerating here some, but the point is that the domain experts are not the sameAs the research logicians.

The research group at Stanford might see this "development of a mapping between "logic based ontology" and XML web services as where the cutting edge is.

So in summary:

Unfortunately our society needs completed results (not long term research projects). Of course, if the long term research project has already have several decades, the funding sources may wish to consider shifting funding so that society's immediate needs are addressed.