John S,
The questions definitely are relevant to the issue of how to use Word
Experts. But the virtue of Word Experts is that the answers can be highly
specific to each word. The method for getting a relevant interpretation can
use any tactic - it is only necessary to represent the result in the terms
and format of the common foundation ontology.
Just one or two specific cases: (01)
> 1. Tom supported the tomato plant with a stick.
. . .
> Making those choices requires quite a bit of background knowledge,
> and it's definitely nontrivial with current technology.
. . .
> For sentence #1, a person who knew something about gardening could
> form a clear image of the relationship between the stick and the
> plant. But somebody who had never seen a tomato plant and had no
> idea whether tomatoes grow on trees like plums or on the ground
> like pumpkins wouldn't really understand that sentence. Such a
> person might be able to guess at the relationships, but could you
> still say that the person "understands" the sentence?
>
Good illustration. If the specific knowledge of how sticks are used as
props for vine plants is not in the ontology, that sentence could not be
understood in the way a teenager would understand. Guessing is not
understanding. Those kinds of pragmatics is what would be included in the
Word Expert, or in a 'topic expert' that would have a broader understanding
of plants and gardening generally, not only tomatoes. The Word Experts will
only understand uses of the word for which it has specific (or hierachically
inherited) knowledge. (02)
> With either way of organizing the knowledge -- by words or
> by subject matter -- how would you relate the lexical info
> about each word to the ontology and to the background
> knowledge about how to play bridge or work in the garden?
>
That is precisely the task of integrating the Word Experts with the
foundation ontology. It is not trivial, and is part of the complexity of
the approach. Nevertheless, at this point I believe that it is necessary.
Perhaps as John Bateman suggests, his implementation of CCG will serve as
well. I hope we shall soon be exploring that question. (03)
>
> Is it likely that a bunch of people (similar to the Wikipedians)
> would be willing and able to enter the kinds of knowledge in the
> kinds of formats necessary for a system that understands?
>
Some, at least, and perhaps a lot. There were experiments to make such
knowledge entry more attractive by structuring it as a game. That tactic
has only been explored thus far at a very preliminary level, and could
probably be pushed much farther. The MIT ' Open Mind Common Sense' project
was disrupted by the death of Push Singh. Doug Lenat's vision was to get
some level of language interpretation, and use that to improve the ability
to learn more by being told. There was a demo of that Cyc learning method
circulated in 2000. But even now I can't find it on the web. Doug keeps
saying that he is only slightly behind his original schedule. Perhaps one
could get an experimental NLU function with a learning component into one of
the on-line virtual reality games, synergistic for the game developer and NL
researcher. (04)
> The Cyc project has been paying professional knowledge engineers
> to enter such knowledge into their system for the past 22 years.
> They had two million axioms in 2004, but Cyc still can't read
> a book in order to build up its knowledge base. How much more
> would be needed? Or is there some better way? What way?
There is no way to know from the outside, for a complex system, when it
doesn't perform as hoped, what part of the development process is
responsible. I would guess that either:
(1) there just wasn't enough support for the basic linguistic effort; what
there was focused one 'practical' topics like business activities, terrorist
activities, military operations, or specialized chemistry knowledge. When I
saw a demo a couple of years ago, it appeared to me that there had been
little serious effort to get deep language understanding of the most basic
concepts, before tackling specialized topics. I believe that without first
getting a deep understanding of the basic part of the language (the defining
vocabulary), efforts to interpret more specialized topics are doomed to
produce shallow analysis, regardless of the capability of the supporting
ontology. Or
(2) The language interpretation project was just not structured properly.
As best I can tell, Cyc's language interpreter does not use Word Experts.
Big mistake. ;-) That is not to say it isn't 'state of the art'. The
problem is that the state the art of NLU is not enough for human-level
interpretation, ontology or not. Or maybe they didn't use CCG, which may be
sufficient, if JB is right. (05)
Pat (06)
Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx (07)
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
> Sent: Sunday, March 30, 2008 11:26 PM
> To: [ontolog-forum]
> Subject: Re: [ontolog-forum] What is "understanding" - was: Building on
> common ground
>
> Pat C. and John B.,
>
> I'll accept parts of what both of you are saying, but with
> many qualifications. Before getting to the qualifications,
> I'll quote an example I've used before.
>
> Following are four sentences that use the same verb in
> a similar syntactic pattern, but with very different,
> highly domain-dependent senses:
>
> 1. Tom supported the tomato plant with a stick.
> 2. Tom supported his daughter with $20,000 per year.
> 3. Tom supported his father with a decisive argument.
> 4. Tom supported his partner with a bid of 3 spades.
>
> Each sentence has the same basic syntax:
>
> NP1(name Tom) V(supported) NP2 Prep(with) NP3
>
> If you used a dependency parser with thematic (or case)
> labels, NP1 represents the agent of the verb 'support',
> NP2 represents the theme, and NP3 represents the instrument.
>
> If we look at a typical dictionary (say the Merriam-Webster
> 7th Collegiate), we find 6 primary word senses, each with
> 1 to 6 subsenses. Without using a computer, I'd pick the
> following senses for each of the four sentences:
>
> 1. Sense 4a: "to hold up or serve as a foundation or prop for"
>
> 2. Sense 3: "to pay the costs of : maintain"
>
> 3. Sense 2a2: "to uphold or defend as valid or right"
>
> 4. Sense 2b3: "to bid as in bridge to show support for"
>
> Making those choices requires quite a bit of background knowledge,
> and it's definitely nontrivial with current technology. But the
> next question is what to do with that choice. It might be useful
> in machine translation for picking the correct verb in some target
> language. Perhaps a statistical translator with enough data could
> do so. But could that be called "understanding"?
>
> For sentence #1, a person who knew something about gardening could
> form a clear image of the relationship between the stick and the
> plant. But somebody who had never seen a tomato plant and had no
> idea whether tomatoes grow on trees like plums or on the ground
> like pumpkins wouldn't really understand that sentence. Such a
> person might be able to guess at the relationships, but could you
> still say that the person "understands" the sentence?
>
> What would sentence #4 mean to somebody who didn't play bridge?
> Or to somebody who had no knowledge of card games of any kind?
>
> Suppose we had a word-expert analyzer with an enormous amount
> of information about each verb. Would that be the best way to
> organize the knowledge base? Would you put some knowledge about
> bridge or tomatoes into some rules for each verb, noun, and
> adjective that might refer to bridge or to tomatoes? Or would
> it be better to put all the knowledge about bridge in a module
> that deals with bridge and all the knowledge about tomatoes in
> a module that deals with tomatoes?
>
> With either way of organizing the knowledge -- by words or
> by subject matter -- how would you relate the lexical info
> about each word to the ontology and to the background
> knowledge about how to play bridge or work in the garden?
>
> If you intend to use logic, how much logic would be needed
> for those sentences? What would a theorem prover do to aid
> understanding? Could proving some theorem about tomatoes be
> considered understanding? Would forming a mental image (or
> a virtual reality image) be closer to understanding? How
> would such images be related to the lexicon and the ontology?
>
> Is it likely that a bunch of people (similar to the Wikipedians)
> would be willing and able to enter the kinds of knowledge in the
> kinds of formats necessary for a system that understands?
>
> The Cyc project has been paying professional knowledge engineers
> to enter such knowledge into their system for the past 22 years.
> They had two million axioms in 2004, but Cyc still can't read
> a book in order to build up its knowledge base. How much more
> would be needed? Or is there some better way? What way?
>
> John Sowa
>
>
>
> _________________________________________________________________
> 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
> (08)
_________________________________________________________________
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 (09)
|