ontolog-forum
[Top] [All Lists]

Re: language vs. ontology was Re: [ontolog-forum] April 20 session on t

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Jack Park <jackpark@xxxxxxxxxxxxxx>
Date: Sat, 08 Apr 2006 17:28:10 -0700
Message-id: <4438551A.3070502@xxxxxxxxxxxxxx>
Below, we see this comment:
> For automatic processing that has real-world effects, the system is
> "told" the meaning by associating actions with specific states    (01)

Further below, we see this comment:
> But whether or not a cognitive system can properly
> assign "meaning" to inputs must be judged by whether the system
> responds appropriately to the inputs, according to whatever goal or
> motivational schema that cognitive system is using.    (02)

Who, in the context of the Durusau comments, makes such a judgment? A 
machine? I don't think so. Not in that context. I suspect, again, in 
that context, that the "telling" of meaning (ontology engineering) must 
be grounded in known ontological agreements among those humans involved.    (03)

The quote:
> The issue of how to "ground" the meanings of symbols is a
> lot more complex than just saying that a person is going to read it
> before any meaning is assigned.  If that were the case, no system could
> ever hope to perform automatic processing, even for something as
> trivial as sending a late-payment notice, without human intervention.    (04)

in my read, seems out of context. Crafting an ontology that drives a 
billing system must always be grounded in an expectation that whichever 
vocabulary the billing system's users apply must be reflected in the 
behaviors of the system.  I don't think Patrick Durusau's comments 
suggested otherwise. I am not taking a position on the ages-old 
arguments, just having trouble following this thread.    (05)

Jack    (06)

Cassidy, Patrick J. wrote:
> Re: meaning of symbols -
> 
> Patrick Durusau wrote:
> 
>>>The "meaning" of a symbol is always determined with recourse to
> 
> human 
> interpretation.
> 
> Not exactly.  The issue of how to "ground" the meanings of symbols is a
> lot more complex than just saying that a person is going to read it
> before any meaning is assigned.  If that were the case, no system could
> ever hope to perform automatic processing, even for something as
> trivial as sending a late-payment notice, without human intervention.
> For automatic processing that has real-world effects, the system is
> "told" the meaning by associating actions with specific states (an "if
> state, then action" rule).  In the case where the ontology has only
> state descriptions, the action rules are typically part of the
> procedural code that uses the ontology, but they could also be part of
> an executable specification.
> 
> The meaning of "meaning" is a much-debated topic, but properly
> constructed, an automatic system can tell whether or not some set of
> symbols (e.g. words in a sentence) are in fact meaningful, or are
> gibberish - according to whatever set of rules the system is taught,
> including semantic and pragmatic, not just syntactic rules.  If the
> system is able to understand and respond to language with the facility
> of a human, it can be fairly said to have understood the meaning of the
> sentences it responds to properly.  No current system can do that at
> even the level of a six-year-old, but there are systems that do some
> interpretation to a certain extent, and they can be fairly said to have
> understood some part of the meaning.  Meaning isn't usually an
> all-or-nothing measure, but graded.  An expert can usually understand
> more of the meaning of a technical description than a novice can.
> 
> There are various ways to "ground" the meaning of symbols.  One is for
> a symbol-processing system to be able to recognize instances of classes
> in the ontology.  At a simple level, a computer reading a file on a
> disk and tokenizing the sentences and recognizing the legal "words" in
> the file has grasped part of the meaning of that file, a real-life
> physical object.   For a computer to do all the kinds of things
> college-educated people do, a lot more of the meaning has to be
> recognized, but getting to that point is part of our current task, no?
> 
> Those working in robotics have to be very careful to assign the proper
> "meaning" to the symbols their programs use, or the robot fails to do
> its task, in a very visible manner.  Recognizing things in the real
> world and responding appropriately is to me very clear evidence that
> some elements of "meaning" of the input signals have been properly
> interpreted - without a person intervening at any point from receipt of
> the "symbol" (sensor inputs) to response.
> 
> Perhaps you want to argue that any elements of "meaning" that a
> computer may seem to understand must first be put there by people.  Of
> course, machines are artifacts created by people.  (Yawn)  People are
> objects created by evolution.  The fact that it took millions of years
> for humans to develop a way to learn how to assign meaning to their
> linguistic inputs, and takes years for individuals to learn how to
> assign meanings to linguistic inputs, but only seconds or minutes to
> upload a complex cognitive system into a computer, is certainly of
> great significance.  But whether or not a cognitive system can properly
> assign "meaning" to inputs must be judged by whether the system
> responds appropriately to the inputs, according to whatever goal or
> motivational schema that cognitive system is using.  The adequacy of
> the response is a different issue from how the system learned to make
> that response.
> 
> I think it is quite appropriate to say that some existing
> knowledge-based programs can assign some elements of meaning to the
> inputs (symbols) they get, and respond appropriately, without any human
> interpretation of the input.  In most cases, it is nowhere near as much
> meaning as people can currently assign, but "meaning" nevertheless.
> 
> Pat
> 
> Patrick Cassidy
> MITRE Corporation
> 260 Industrial Way
> Eatontown, NJ 07724
> Mail Stop: MNJE
> Phone: 732-578-6340
> Cell: 908-565-4053
> Fax: 732-578-6012
> Email: pcassidy@xxxxxxxxx
> 
> 
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
> [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Patrick
> Durusau
> Sent: Saturday, April 08, 2006 1:01 PM
> To: Adam Pease
> Cc: [ontolog-forum]
> Subject: Re: language vs. ontology was Re: [ontolog-forum] April 20
> session on tagging ontolog content
> 
> Adam,
> 
> Adam Pease wrote:
> 
> 
>>Patrick,
>>  I think there's a misunderstanding here.  It's not just a question 
>>of precision in the conventional sense.  Of course we try to make our
> 
> 
>>documents and laws precise.  But the meaning of those texts is 
>>determined with recourse to human interpretation.  In a formal 
>>ontology like SUMO, or DOLCE, one could replace all the term names 
>>with arbitrary unique symbols, and an automated deductive system, 
>>following the rules of mathematical logic (in our case, first order 
>>logic) could reach all the same conclusions as it could if those 
>>intelligible labels were present.  The meaning of the symbols is 
>>defined mathematically, and no human interpretation is required to 
>>give them meaning.
> 
> 
> Yes, there is a misunderstanding here.
> 
> defining a mathematical logic between arbitrary unique symbols != 
> defining the meaning of arbitrary unique symbols
> 
> Yes, whatever symbols are substituted, the outcomes of defined 
> operations would be the same.
> 
> That is not the same as defining the "meaning" of a symbol
> mathematically.
> 
> The "meaning" of a symbol is always determined with recourse to human 
> interpretation. How else would you explain the use of WordNet
> references 
> in SUMO? Of what possible relevance are WordNet entries if the meaning 
> of symbols is being defined mathematically? The answer is that the 
> "meaning" of the symbols is being defined by making references to 
> something outside of the mathematical definition of relationships 
> between the symbols.
> 
> 
>>  That's relevant because it enables one to do things like prove the 
>>absence of contradictions in the use of these terms with an automated
> 
> 
>>system.  In contrast, without such a property, humans have to 
>>determine whether usage of terminological or linguistic based 
>>standards are compliant.  Standards compliance is something I would 
>>think would be relevant to the world bank.
>>
> 
> Not to put too fine a point on it but you seem to be eliding over the 
> fact that the development of SUMO was based upon human interpretation 
> and construction of the rules that can then be automatically applied.
> It 
> wasn't simply a matter of putting mathematical rules in a box of
> symbols 
> and shaking it until the result suddenly appeared.
> 
> The compliance the world bank (or any other concern) is interested in
> is 
> compliance with their terms, which may or may not use the same rules
> for 
> complaince that can be found in SUMO. Yes, logic can prove a lack of 
> contradiction but then so can careful definition of terms and what is
> or 
> is not considered to be a conflict.
> 
> Note that I do consider SUMO and similar efforts to be worthwhile. But,
> 
> I don't think that they divorce the meaning of symbols from human 
> interpretation nor are they the best solutions for all cases. As I
> noted 
> earlier today, we have been down the road of universal languages before
> 
> and none has succeeded to date.
> 
> Hope you are having a great day!
> 
> Patrick
> 
> 
>>Adam
>>
>>At 03:08 AM 4/8/2006, Patrick Durusau wrote:
>>
>>
>>>Adam,
>>>
>>>Adam Pease wrote:
>>>
>>>
>>>>Folks,
>>>>  For what it's worth, I think there is a common problem surfacing 
>>>>here, that Bill has tried to point out.  Language and ontology are 
>>>>different.  Human language (and any given word in a human language)
> 
> 
>>>>is ambiguous and highly contextual.  Terms in an ontology are not 
>>>>ambiguous (or at least, shouldn't be if they are properly and 
>>>>formally defined).
>>>>  Typically, this has been a problem, because computational 
>>>>linguists have often used linguistic elements as pseudo-logical 
>>>>terms in semantic forms.  Ontology builders often use linguistic 
>>>>elements as proxies for doing a full semantic definition, leaving 
>>>>much of the interpretation embedded in the conventional meaning of 
>>>>the linguistic-based term.
>>>>  The approach we've taken in SUMO is to make this distinction 
>>>>explicit, and to address language and ontology in separate but 
>>>>related products.  SUMO is the formal ontology with terms defined 
>>>>unambiguously in first order logic.  Those terms are related
> 
> through 
> 
>>>>semi-formal links to the word senses in Princeton's WordNet.
>>>
>>>
>>>Language and ontology are different???
>>>
>>>Hmmm, well the foundational paper for SUMO states:
>>>
>>>"In order to enable continued progress in ecommerce and software 
>>>integration, we must give
>>>computers a common language with a richness that more closely 
>>>approaches that of human
>>>language."
> 
> http://home.earthlink.net/~adampease/professional/FOIS.pdf
> 
>>>Granted a great deal of effort has gone into making SUMO precise,
> 
> but 
> 
>>>the same could be done for any language. It is interesting but not 
>>>persuasive that its terms have been "defined unambigouously in first
> 
> 
>>>order logic." And that is relevant for what reason? Perhaps first 
>>>order logic is not relevant to all the problems faced by the World 
>>>Bank. Recall that the current fascination with first order logic is
> 
> a 
> 
>>>repeat of a debate that has ebbed and flowed for many years. Justice
> 
> 
>>>Holmes wrote in the 1890's that the life of the law had been 
>>>experience and not logic. 
>>>(http://www.gutenberg.org/files/2373/2373-h/2373-h.htm)
>>>
>>>In any event, there is no reason to disenfranchise the World Bank 
>>>from representing their language/ontology in favor of using SUMO. 
>>>There have been any number of attempts to produce universal 
>>>languages, LogLang is one of the more recent ones.
>>>
>>>There are standards that seek to empower users to define their own 
>>>languages/ontologies and yet remain mappable to others. See, for 
>>>example the Topic Maps Reference Model CD draft at: 
>>>http://www.jtc1sc34.org/repository/0710.pdf.
>>>
>>>Hope you are having a great day!
>>>
>>>Patrick
>>>    (07)

_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Unsubscribe/Config: 
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (08)

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