[Top] [All Lists]

Re: [ontolog-forum] Universal Basic Semantic Structures

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Wed, 05 Sep 2012 15:54:32 -0400
Message-id: <5047ADF8.8020203@xxxxxxxxxxx>
Kingsley, Andries, Doug,    (01)

> What's not to like about this excerpt:    (02)

> In its simplest form, this is a structure that is also supported by 
> such as _RDF_ and _OWL_. However, a semantic model includes the following 
> extensions that support an improved computer interpretation of such sentences
> and an improved computerized verification of semantic correctness:    (03)

The goals are important, and I believe that Andries has put together
an impressive system of interrelated ideas.  I believe that Gellish
contains a lot of good material that is worth considering.    (04)

My major concern is that it is YASS -- Yet Another Semantic Structure
that is designed to solve all the world's problems -- but ONLY IF the
entire world swallows it whole and converts everything to it overnight.    (05)

The biggest YASS of all is Cyc, which had one person millennium of work
invested in it as of 2010.  Cyc is very impressive.  It has implemented
many things very well, and Doug is better able to explain those than I
am.  But as we have seen, it still has not been swallowed whole.    (06)

The Semantic Web is a cooperative project to build YASS, but it is
now 18 years since Tim B-L's original talk on the subject, 12 years
since his DAML proposal, and 6 years since the DAML final results.
But the uptake of the SW is still glacially slow.  If you add all
the work spent on the SW, it's probably quite a few person millennia.    (07)

Meanwhile, much less tightly organized things such as Linked Open Data
have had a much faster uptake with a much looser organization.  The
question of how to combine ideas from LOD with any of the many YASSes
is still an unanswered research issue.    (08)

> Originally Gellish was developed as a generic data model. Data modelers are
> generally not aware that a data model is in fact a kind of primitive
> language definition and that that is the cause of the huge data integration
> problems. That is also the reason why it took time before Gellish developed
> in the direction of CNL's.    (09)

The orientation toward practical applications of software engineering
is important.  My major criticism of Cyc was that they did not address
applications in the early days.  They have been devoting much more
effort to applications recently.  I believe that has had an important
effect of focusing attention on what works.    (010)

> Where I differ from Doug is that I am looking at the basic foundation
> of factual expression being a 3-tuple or 4-tuple (but this lacks
> standardization at this time, for instance we do things in our product
> that's different from others re. role of the 4th element).    (011)

This illustrates the weakness of mixing logic and ontology without
having a clear idea of either one.  RDF is basically a very simple
version of logic.  The fourth element was added to solve some issues
that belong to ontology, but nobody had a clear idea of what those
issues are or how they relate to the logic.    (012)

> Quads allow a pointer to be attached to a triple.  That pointer can reference
> an object that is split into context and provenance (or equivalently a
> context which has specified provenance).  Quads do not allow for higher
> arity relations -- except for ternary relations if context and provenance
> are ignored.
> And yes, i know that a module for provenance has recently been patched
> onto RDF.    (013)

Provenance is important.  But it is a metalevel notion, and you need
a general metalevel mechanism to address it.  In principle, provenance
is every bit as complex as any issue in ontology or any other subject.    (014)

There is no way that you can adequately handle provenance or any other
metalevel issue just by tacking an extra field on each triple.  That
extra field is just going to be misused in hopelessly confused and
"creative" ways.  It's just an open invitation to disaster.    (015)

> I am indeed convinced that free n-tuples are a pain, because everybody is
> free to create his own n-tuples and there is no common convention to
> interpret them. The relations between the elements in the tuples are not
> explicit, neither are the roles that those elements play in those relations.
> My research is oriented towards a common formal language and a universal
> data structure.    (016)

I strongly agree with Andries on this point.  I believe that the SW
tacked an extra node on the RDF triples without doing that research.
But I still have reservations about the possibility of any ontology
being truly "universal".    (017)

> I am simply fighting for a starting point that maps closely to
> mental model of most humans.    (018)

Human mental models are based on imagery.  Their relationship to logic,
cognitive science, and neuroscience is still a major research area.
See the following slides, in which I discussed those issues and their
relationships to Peirce's existential graphs (EGs):    (019)

    "Moving pictures of thought" -- Peirce's existential graphs    (020)

EGs are significant, especially since they are upward compatible
with RDF and they have a solid mathematical foundation.  But they
illustrate the enormous number of still unsolved R & D issues.    (021)

Summary:  Many very intelligent people have invested many person
centuries (or millennia) of very hard work on these issues.  But
I don't see any hope of convergence on these issues any time soon.    (022)

Meanwhile, I'd like to point out the following slides, which
discuss the problems of legacy software.  None of the above
efforts address those problems.    (023)

___________________________________________________________________    (024)

Excerpts from http://www.jfsowa.com/talks/iss.pdf    (025)

                  What are the Problems?    (026)

Semantics should facilitate the interoperability and
integration of all systems – legacy, current, and future.    (027)

But there is no such thing as a one-size-fits-all set of
languages and tools for processing semantics.    (028)

People with different jobs and skills require different tools:    (029)

   ● Most semantic tools and languages are too difficult for the average
     programmer, website developer, or subject matter expert (SME).
   ● They are also too limited for people who are capable of designing
     better tools for programmers, website developers, and SMEs.
   ● Different kinds of applications require different amounts of detail:
     lightweight, middleweight, or heavyweight semantics.
___________________________________________________________________    (030)

                    Size of the Problems    (031)

Some estimates:    (032)

   ● World-wide digital data in 2009: 800 million terabytes.
   ● Estimated digital data in 2010: 1.2 billion terabytes.
   ● Legacy software: Half a trillion lines of code.
   ● Percentage that is tagged with semantics: Slightly over 0%
___________________________________________________________________    (033)

                  How Could Semantics Help?    (034)

Typical programmer productivity with current tools:    (035)

   ● 10 to 15 lines of fully debugged code per person per day.
   ● Cost per line of code: $18 to $45.
   ● Most of the time is spent on analysis, design, and testing.
   ● Specification errors are the most costly and time consuming
     and the most likely to benefit from clearly defined semantics.    (036)

Why semantic technology isn’t used more widely:    (037)

   ● Too much time, effort, and training to specify semantics.
   ● Delayed implementation without any obvious benefit.
   ● Difficulty of sharing semantics among different tools,
     especially tools designed for different methodologies.    (038)

We need better tools and better integration among tools.    (039)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (040)

<Prev in Thread] Current Thread [Next in Thread>