SUO: Semantic interoperability
At 10:10 2001-12-08 -0500, John F. Sowa wrote:
> Frank,
>
> When I mentioned logic, I intended it in a broader sense than
> just a "machine representation":
>
> Frank Farance wrote:
>
> > Here was my note on the NCITS/L8 in response to Joe Christensen. In
> > the usage below (which, I believe, serves Joe's purposes), agreement
> > on "meaning" is not necessarily dependent upon a machine representation
> > of meaning.
>
> Logic was developed by Aristotle long before there was any machine
> that could process it. By logic, he and his followers meant a precise,
> highly stylized version of controlled natural language. The symbolic
> logic that was invented in the 19th century used a more mathematical
> notation, but it was also intended to be written and read by humans.
Sounds like you agree.
> In the example I gave, I showed how you could take a vague statement
> in English and replace it with a precise statement, also in English,
> that could, if one wished, translate directly to a formal notation
> such as CGs or KIF. All that could be done independent of any machine
> or software tools that might process the representation.
Yes, but more precision isn't always better (especially for standards). <-- FYI, this comment might not apply to the SUO WG, but could easily apply to other standards committees.
> > The ISO/IEC 2382 Information Technology Vocabulary (ITV) might be able to help you. The definition of "interoperability" is:
> >
> > 01.01.47 interoperability: The capability to communicate, execute programs, or transfer data among various functional units in a manner that requires the user to have little or no knowledge of the unique characteristics of those units.
> >
> > In short, "interoperability" means communication/ execution/ data transfer without knowing the nature of the implementations (e.g., the endpoints of communication, the execution environment, data repositories, etc.).
>
> Both the ISO definition and the paraphrase are examples of what I
> called "big fuzzy clouds" that sound good, but are too vague to support
> systems that can actually "interoperate" -- i.e., process the same
> data in logically equivalent ways.
>
> I don't deny that those statements are useful starting points, but
> a standard for interoperability must be stated in much more precise
> terms to be useful.
>
> > I got the following definition from "dictionary.com":
> >
> > se·man·tic (s-mntk) also se·man·ti·cal (-t-kl) adj.
> >
> > 1. Of or relating to meaning, especially meaning in language.
> > 2. Of, relating to, or according to the science of semantics.
> >
> > I believe the construction "semantic" [dictionary.com] + "interoperability" [ISO/IEC 2382] would work well:
>
> This is another big fuzzy cloud of words. It defines semantic in
> terms of meaning, which is itself defined in terms of semantics.
> Unless you have a good intuitive feeling for the subject, it doesn't
> help in any way.
>
> > semantic + interoperability: In the conext of or relating to *meaning*, the capability to communicate, execute programs, or transfer data among various functional units in a manner that requires the user to have little or no knowledge of the unique characteristics of those units.
> >
> > or shortened a bit:
> >
> > semantic interoperability: The capability to preserve meaning in communications, program execution, and data transfer among various functional units.
>
> Again, this doesn't define meaning or give any criteria for what
> it means to preserve meaning.
>
> > In other words, the above definition should be compatible with "interoperability" when used in standards wording.
>
> Again, sprinking a few words like "shall" into the mix might make
> it look like "standards wording", but it won't make the definition
> precise enough to enable two systems to operate in the same way on
> the same data.
I think you confuse "sprinking a few words like "shall"" with standards wording. The point is that humans (especially standards folk) have been able to interpret terms "interoperability" just fine. In other words, standards wording is defined only as precise as necessary ... there are good reasons (both theoretical and practical) for limiting the precision ... this "fuzziness", as you describe, can be useful. Certainly, too much precision can be problematic for certain applications.
You are correct that I did not define "meaning" in my suggested definition. However, the "meaning" comes from the standards (or specifications) that describe the interoperability. Some standards describe lots of "meaning", while others describe (sometimes, purposely) little "meaning". Putting all the pieces together (i.e., the compilation of standards used in, say, data interchange) then gives one an idea of how much "meaning" is associated with the data interchange and, how much "semantic interoperability" exists or is possible.
As I've said, many standards committees have been able to interpret the term "interoperability" reasonably consistently (within a committee), even if it means different things for different committees (and it should) ... similar to the term "quality" ... it is difficult to use precisely without context.
-FF
_______________________________________________________________________
Frank Farance, Farance Inc. T: +1 212 486 4700 F: +1 212 759 1605
mailto:frank@farance.com http://farance.com
Standards, products, services for the Global Information Infrastructure