Re: [ontology-summit] The tools are not the problem (yet)

["Barkmeyer, Edward J" <edward.barkmeyer@xxxxxxxx>]

John,

 

I agree with much of this.  As John Sowa is fond of saying, humans can work very well with fuzzy classification.  The problem for knowledge engineering is that software (as we know it) cannot.  And when you trust a decision to software, you need to know what its classification rules are.

 

One view of knowledge engineering and inference is that it is primarily a means of filtering data to provide human arbiters with the best information for making a decision.  In that case, your knowledge engineering is not significantly harmed by fuzzy terms and occasional ‘misuses’.  Our application of knowledge engineering is to provide automata with an ontology for the work and the workspace that enables them to make real-time decisions (without consulting humans).  In that case, a fuzzy term means you don’t know what the automaton will do in some situations, and from a business point of view, that can be very dangerous.  (Think of the unexpected interpretations of keystroke-in-context in various tools.)

 

If I understand your argument below (and I doubt that), (c) means that the conceptually integral predicator, i.e., “relation” (?), is often fragmented into multiple RDF predicates.  To do (d) you need to know what the intended relation is, and what the fragmentation pattern is.  (That is exactly what is going on for one of the ISO 15926 methodologies.)  And if you have no way of knowing that, you get a 1-to-1 translation, which involves essentially meaningless predicates.  But I don’t see how that makes predicates the problem.  The problem seems to me to lie in the choice of a syntax that doesn’t allow the intended relation to be expressed as a predicate. 

 

At another level, however, the meaning of a FOL _expression_ is exactly what the model theory says it means, no more and no less.  Whatever you name as a relation is one.  What it really means to the knowledge engineer is beyond the concern of the FOL itself.  Some FOL dialects make it more or less difficult for you to express as a “predicate” what you intend to be an integral ‘relation’ that describes a ‘state of affairs’, if you like.  But philosophically, that is a can of worms.  Who is to say that your perception of what is integral is really ‘integral’ in any fundamental way?  In that view, all predicates are artificial. 

 

Put another way, mathematical logic, like all of mathematics, is intrinsically a science of the artificial.  It just turns out that some mathematical models of perceived reality yield useful information for the person/organization actually functioning in the real world.  And that said, statistical analysis models and logical analysis models are just different kinds of mathematical models, and they can both become “practically unimplementable” when the perceived world is too big and too complicated.  If that is what you mean by “the problem is in the predicates”, then, yes, I agree.

 

The late Dr. James Albus referred to the AI problem as “focusing attention” (which humans do naturally) – somehow filtering the flood of incoming information to that which is useful to solving the problem at hand with the available “mental” tools.  If you are simply overwhelmed by the information flood, you are a “deer in the headlights”, incapable of decision.  If your available tools mis-classify the situation, you may make a bad decision, but if they produce a useful classification, you make an adequate decision.  In that view, the problem is not the plethora of predicates; it is the absence of effective filters for some problem scope.

 

-Ed

 

 

 

From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John McClure
Sent: Thursday, January 23, 2014 4:30 PM
To: ontology-summit@xxxxxxxxxxxxxxxx
Subject: Re: [ontology-summit] The tools are not the problem (yet)

 

Hi Ed,
I agree much of what you say, but would add that there's 100% certainty that your understanding of Concept:A is differs from my understanding. A 'close enough' practical view of reuse is what drives my approach in this matter. In fact many ontologists (and programmers!) mechanistically subclass/subproperty each item in a reused component, specifically to allow for specializations (interesting that there's no subindividuation possible that I know of).

Getting to your defense of Doug's view that RDF is so lacking it's useless for functionally distinguishing between concepts, I am still replying that that's irrelevant to my critique. Let me restate my critique here.

• (a) If FOL representation is the ideal, then RDF as an FOL subset should conform with its rules
• (b) The rule for FOL representation is that FOL predicates are coordinate with predicators
• (c) RDF properties are -- far far too often -- not conforming to this rule for FOL predicates
• (d) RDF predicates (during ingestion) must be transformed to FOL predicates when non-conforming
• (e) When not so transformed, FOL predicates are being created that break FOL rules

Consequently the sets of FOL and RDF predicates have exploded to unuseable sizes.

Again, the tools are not the problem -- predicates are the problem. In particular, nounal predicates which flagrantly break the rule for FOL predicates.
/jmc

On 1/23/2014 11:33 AM, Barkmeyer, Edward J wrote:

John,

I beg to differ.  It is not just about how we name things; it is about how we know what is being named.  

A useful taxonomy is based on well-defined properties.  So the capabilities of the chosen syntax make a difference.  Can you phrase:  "An X is a Y that does (or does not) have property P" in the syntax you choose?  Alternatively, can you phrase:  For all X's, sentence S1 (about X's) is true?.  These are the established structures for formal definition.  Ultimately, there must be some terms that are "primitive", in that we can't define them formally in terms of other concepts or axioms.  But if all or most of your terms are primitive, no one can be really sure what any of them mean.

I think it is quite reasonable to consider whether you want to use someone else's formal (e.g., RDF) term for something that the source defines loosely in English from his/her particular point of view.  You have to determine that the definition is unambiguous, in terms of the definitions of the English words used and the syntactic (and pragmatic) context of their use.  Then you have to determine whether that is exactly what you mean for your particular purposes.

For example, suppose we define the IRI http://mydictionary.com#food as "anything people eat".  Why "people"?  (Why do we call it "dog food" -- is that a misnomer?)  Which people did the author have in mind?  Does that include grasshoppers?  Does it include grass?  Do you really want to use his term, thereby implying that anything anyone says about ...#food is consistent with what you say about ...#food?  That is the problem.

The ISO TC37 "terminology" folk would say that agreement to use a set of terms consistently creates a "speech community" who have a "shared understanding" of the meaning of the terms, insofar as they are used within that community.  So "reuse" is a property of terms WITHIN a community.  The problem with RDF (and the Semantic Web in general) is that construction/development of those communities is left as an undefined social exercise.

Doug's point is that it is very difficult in RDF (mostly impossible) to say what you do and do not mean by ...#food.  As a consequence, it is very risky for me to assume that what you mean and what I mean are (exactly) the same, unless there is some community of use that I know we both belong to (and we keep out, disparage, vilify the riffraff who misuse OUR terms).   There are some thriving speech communities using RDF, but their vocabularies are not very large, because the language itself doesn't much help in creating the shared understanding.  OWL is more expressive than RDF, and you can write (some) formal definitions and axioms in OWL, but it takes real dedication to do that wherever possible and get it right.   So, OWL has a combination of RDF-like communities, who formalized the terms and develop the common understanding by informal means, and a few communities whose terms are rigorously defined, and a few communities who aspire to the latter and are still learning how

  to do it.

The point is that these "reuse" communities arise by choosing a language (syntax) and a set of practices that they all agree to be comfortable with.  The shared understanding is the "content", but the means of _expression_ is part of the shared practice that begets and maintains the shared understanding.  You can separate content from syntax in your own head, but you cannot separate content from syntax when you communicate with others.  Syntax, of some kind(s), is critical to creating the "shared understanding" of a term.

-Ed

--

Edward J. Barkmeyer                     Email: edbark@xxxxxxxx

National Institute of Standards & Technology

Systems Integration Division

100 Bureau Drive, Stop 8263             Work:   +1 301-975-3528

Gaithersburg, MD 20899-8263             Mobile: +1 240-672-5800

"The opinions expressed above do not reflect consensus of NIST, 

 and have not been reviewed by any Government authority."

-----Original Message-----

From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-

summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John McClure

Sent: Thursday, January 23, 2014 12:04 PM

To: ontology-summit@xxxxxxxxxxxxxxxx

Subject: Re: [ontology-summit] The tools are not the problem (yet)

On 1/22/2014 10:47 PM, doug foxvog wrote:

<snip/>

"Reuse" was the promise; empirically it has not been delivered - why not?

Because it was expressed in RDF.

Let's not confuse format with content.

 + Class taxonomies + Property taxonomies

If you hate CycL, pick another format. -- doug

Format is not the issue. It is how we name things.

__________________________________________________________

_______

Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/

Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-

summit/

Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx

Community Files: http://ontolog.cim3.net/file/work/OntologySummit2014/

Community Wiki: http://ontolog.cim3.net/cgi-

bin/wiki.pl?OntologySummit2014

Community Portal: http://ontolog.cim3.net/wiki/

_________________________________________________________________

Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/   

Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  

Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx

Community Files: http://ontolog.cim3.net/file/work/OntologySummit2014/

Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2014  

Community Portal: http://ontolog.cim3.net/wiki/ 

 

_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/   
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2014/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2014  
Community Portal: http://ontolog.cim3.net/wiki/     (01)