ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Levels

To: "Deborah MacPherson" <debmacp@xxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 21 Feb 2007 17:31:02 -0600
Message-id: <p0623090fc20282eca1fe@[10.100.0.26]>
>Fascinating. Thanks for all the definitions.
>
>Wouldn't linking background knowledge together eventually simplify
>what actually are persistent ambiguities versus problems that may
>already be addressed by others, or have more concise definitions in
>other languages? (IE no one-word English translation for the German
>word 'gestalt')    (01)

That is one of the ambitions, yes. Actually doing 
it turns out to be almost unbelievably difficult, 
however. The extent and nature of all this 
'common sense' or 'background' knowledge is 
staggering. Just listing it all is something we 
don't really know how to do, and then encoding it 
in a form usable by computers is a research 
problem. In fact, the entire field of formal 
ontology construction emerged from the subfield 
of AI (called 'knowledge representation' or KR) 
which set out to do just this, i.e. to formalize 
human intuitive knowledge in a machine-usable 
form.    (02)

>Instead of languages, take an open question in the field of physics
>for example. Today there are so many specialized areas of physics and
>these methods of thinking, working, communicating, documenting and
>exchanging ideas and information are practiced all over the world. But
>the further back in time you go, less was known, there weren't as many
>labs or publications, or literally as many words and equations to keep
>track of or provide. At some point, every background knowledge trail
>will end because a particular ambiguity had not been imagined or
>recorded yet. Either a person or computer, or network of people and
>machines, notices a new ambiguity and they are already limited in
>saying where the ambiguity belongs in terms of background knowledge
>because the whole issue is new. Words may have been invented to
>describe it, this act itself tells a story.
>
>But I imagine your objective is not to make a clear picture of what we
>knew before, reinvent citation methods, or even show how we advanced
>from star gazing to modern astronomy - but to connect enough accurate
>knowledge being exchanged via computers and networks to see - oh look,
>this ambiguity described by these words is currently being
>passionately argued between XYand Z labs, published in these journals,
>the heart of the matter is expressed by these equations, this is where
>the situation currently stands.    (03)

That WOULD be nice :-) We can't come close to 
doing anything remotely this sophisticated. The 
state of the NLP art currently is, for example, 
being able to order a book on Amazon, given its 
author and title; or having a brief conversation 
with a patient to discover if they have been 
taking their medications on time.    (04)

But you know, most ambiguities aren't observed 
and argued over, they just arise by a kind of 
linguistic speciation which drags meanings apart 
in different subcommunities (e.g. the physicists 
took 'energy' and 'work' into a whole new 
semantic place when thermodynamics required more 
precise ideas.)    (05)

>A machine identifying this ambiguious
>term would then open up these particular floodgates using a mechanical
>version of common sense and collective memory/experience?    (06)

There is no mechanical version of common sense, 
yet. The nearest contender is Cyc, probably the 
world's largest formalized ontology, with 
millions of facts at its disposal; but even Cyc 
is nowhere near what would be needed to mirror 
the 'common sense' of, say, a six-year-old human 
child.    (07)

There is an list of common sense challenge 
problems, very few of which have been so far 
mastered by machines. One I contributed is to 
explain why, when the wolf heard the woodcutters 
singing in the distance, he decided after all not 
to pounce on little red riding hood. My 
granddaughter understood this when she was three; 
no computer system has yet come close to being 
able to, as far as I know. It is quite a 
challenge to write it all out even in English.    (08)

Pat Hayes    (09)


>  Is your
>purpose to limit the *LEVELS* being searched and point towards the
>particular resources involved in the difficulties surrounding
>ambigious terms?
>
>Wouldn't it be great to be fed updates as the ambiguity is removed by
>being solved or getting splintered into solved versus unsolved
>portions of the problem?
>
>Debbie
>
>On 2/21/07, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
>>  Debbie,
>>
>>  Re NLP:  NL is a common abbreviation for natural language,
>>  and NLP is natural language processing.  Many people (including
>>  me) have been working on NLP for years, and we have all learned
>>  that the task is much harder than anyone had anticipated.
>>
>>   > In highly structured documents and words such as a set
>>   > of specifications, don't you think identifying ambiguities
>>   > people do not notice is a different task than looking for
>>   > and highlighting ambiguities that require interpretation?
>>
>>  Yes, they are different tasks.  But the ambiguities that
>>  must be resolved for people (who use "commonsense" to
>>  interpret documents) are just a small subset of the problems
>>  that a computer runs into (because it doesn't have commonsense).
>>
>>   > Is the purpose of auto-nitpicking (what is supposed to be)
>>   > structured text to pull out seemingly minor inconsistencies
>>   > so they can be either dismissed or decided?
>>
>>  If we want to get computers to help us process ordinary
>>  language, we have to provide them with the kind of background
>>  knowledge (i.e., "common sense") that people use.  But we have
>>  to teach them that background knowledge by using more structured
>>  languages that they are capable of interpreting right now --
>>  examples include symbolic logic or programming languages.
>>
>>  The hope is that if we provide enough background knowledge,
>>  then perhaps someday they would be able to read the same
>>  kinds of texts that are designed for humans to read.  Some
>>  people (such as Doug Lenat and the Cyc project) have been
>>  trying to do that for the past 22 years, and they still
>>  haven't succeeded (or Lenat would say that they have only
>>  achieved partial success).
>>
>>  John
>>
>>
>>
>>
>>
>>
>>
>>  _________________________________________________________________
>>  Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
>>  Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>>
>>
>
>
>--
>
>*************************************************
>
>Deborah MacPherson
>www.accuracyandaesthetics.com
>www.deborahmacpherson.com
>
>The content of this email may contain private
>confidential information. Do not forward, copy,
>share, or otherwise distribute without explicit
>written permission from all correspondents.
>
>**************************************************
>
>_________________________________________________________________
>Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/ 
>Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>    (010)


-- 
---------------------------------------------------------------------
IHMC            (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.    (850)202 4416   office
Pensacola                       (850)202 4440   fax
FL 32502                        (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (011)


_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (012)

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