[Top] [All Lists]

Re: [ontolog-forum] Axiomatic ontology

To: Avril Styrman <Avril.Styrman@xxxxxxxxxxx>
Cc: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Fri, 1 Feb 2008 10:51:55 -0600
Message-id: <p06230901c3c8f68092da@[]>
At 4:46 PM +0200 2/1/08, Avril Styrman wrote:
Quoting Pat Hayes <phayes@xxxxxxx>:

> >A large number of simple "X is Y" NL sentences gives a quite
> >good approximation that "X is a subclass of Y"
> OK, if you believe this, find me some. Actual examples from the Web.
6 040  hits "tulip is a flower".
56 200 hits "flower is a plant".
plant > flover > tulip

In course of objectivity, we also get:

67 000 hits "plant is a flower"
4 100  hits "flower is a tulip"
tulip > flover > plant 

The point is, that we do get all these very
easily, and after we have got them, we can
reason about them based on their relations.

You only 'get' them if you already know what you are looking for. (In fact, I even doubt this, as this search will be finding a large number of sentence fragments, whose actual meaning might be almost anything: consider for example "This particular plant is a flower that blooms during the day" (from the first Google page of hits) which inverts the class relationship suggested by the match to the "plant is a flower" search.
If you search instead for patterns of the form 'X is Y', you will get an overwhelming number of instances which do not map into simple class relationships. Now, to establish your case, you need to show me how your system will distinguish one kind from the other.

By the way, your example would seem to suggest that the preponderance of evidence (67K rather than 56K) is that plants are a subclass of flowers, which is clearly nonsense.

> >The main point was the web is more
> >likely to be the repository of common sense than Cyc.
> This is false on many grounds. First, common sense is not commonly
> written down, even in English. Some of it may *never* be written down
> because we all learn it (or are born with it) before becoming
> literate. Second, it is not currently or for the forseeable future
> possible to extract meaning from free NL text without already having
> a CYC-scale ontology to help one comprehend the NL sentences with.
> Third, although there may indeed be a great deal of information on
> the Web, there is no way to distinguish facts from opinion or sense
> from nonsense, unless one already has a great deal of common sense
> (more than many adult humans) to help one do the sorting out.

Imagine that you had the task to build a machine that passes the
Turing-test. Suppose that you had 100 years to do this, and you had
100 experts working for you full-time.

I would not accept this task. The Turing Test is over-rated, and bears almost no relationship to actual AI work. But this is a poor start for you to make your case, as even a human being could not extract common sense from the entire Web.

You would also have all the
current equipment that you need. But you would have to manage with
either 1) Cyc as it is now, of with 2) Internet as it is now, without
Cyc of course. Would you choose Cyc?

Yes, though admittedly with gritted teeth. Its not that I particularly love Cyc, but there is currently no alternative.

If you did, your experts would
be eventually typing the same common-sense knowledge that is
already in the Internet.

It isn't in the Internet.

If you'd choose Interner, you wouldn't have
to type the common sense knowledge that is in Cyc.

Yes, you would. Or at any rate, nobody has come up with any alternative yet.

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

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    (01)

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