John,
Perhaps a little ambiguity on my side:
[JS] > Re reliable sources: If an issue is important, I would never accept
> any answer from anybody without getting a second or third or fourth
> opinion. But many of those sources are also available on the WWW:
>
>> PC> ... like dictionaries, encyclopedias, glossaries, thesauri, and
>> other ontologies, and occasionally textbooks.
>
> I also download many classic books and quite a few articles on my
> own computer so that I have my preferred selection, which I can
> search with Google Desktop.
>
Yes, of course, agreed on both counts. I do too. Almost all the
information I use these days comes from the Web. What I meant was that
accessing Web information without any accounting for the likely reliability
of specific pages is a very risky enterprise, and it requires human-level
language interpretation to determine what is reliable. At present it seems
impossible to use the whole Web (rather than specific sites with understood
formats) as data for anything but a search tool that presents a human with a
list of possible answers. (01)
[Re: quality of automatic translation]
[JS]>
> As for quality, it wasn't a problem because the most important info
> was in the equations,
Yes, and in biochemistry most of the information in many articles is in
tables, and one only needs to translate the column headings for most
purposes. Been there, done that - I occasionally find machine translation
*useful* for the same reasons you mention. But the issue that I was
addressing was whether any form of automatic translation can be interpreted
as "understanding" (a word you used) at anything close to the human level,
and the statistical translation methods don't at present have that
capability, because they are not trained on texts that are marked up with
detailed semantics. Some day, statistical translation may come close, if it
is possible to tag many texts with precise semantic categories and have
those texts mapped to semantic interpretations. I just expect rules-based
language understanding to get there first. But the relative virtues of
statistics and rules for human-level *understanding* will only be discovered
as more work is done. (02)
[Re: spatial reasoning]
> PC> Do you know of any programs publicly available that actually can
> > do the kind of spatial reasoning that people can do? Do you think
> > that virtual reality and other simulations actually reason spatially
> > in some way similar to the way people do?
>
[JS] > Both computers and humans can transform images by rotations,
reversals,
> combinations, etc. But there are still many unsolved questions about
> how the human brain does anything.
>
> In any case, the algorithms we use for computing semantic distance
> measures on graphs (conceptual graphs, by the way, but that really
> isn't much of a restriction, since CGs can represent any graph) can
> be adapted to compute semantic distance measures on geometrical
> representations.
>
> Whether those operations are similar to what goes on inside the brain
> is unlikely, but the crucial test is whether the measures that the
> computer would determine as close or distant have a high correlation
> with what people would consider close or distant.
>
Again, I probably need to add more detail to clarify the question:
Do you know of any programs that can do spatial reasoning so as to be able
to answer the kinds of spatial-reasoning questions one finds on college
entrance exams? Without knowing the mechanism of human reasoning, answering
such questions are examples of what some think of as indicators of
"intelligence". (03)
[JS] > If you can hold your nose to look at Google, you can type "spatial
> reasoning" to get 329,000 things to look at -- or not, as you prefer.
>
Er - I think that makes my case, no? ;-) (04)
Pat (05)
Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx (06)
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
> Sent: Friday, February 01, 2008 11:05 PM
> To: [ontolog-forum]
> Subject: Re: [ontolog-forum] Axiomatic ontology
>
> Pat,
>
> PC> Show me a text that requires a little thought to understand,
> > and then look at an automatic translation. Yechh! "Understanding"
> > it is **not**.
>
> SYSTRAN was used to translate many pages of physics articles from
> Russian to English during the 1960s (after Bar Hillel had said that
> MT would never be practical). But the physicists actually liked
> it -- and they often *preferred* MT to a human translation.
>
> They reasons they gave were interesting: They could get a machine
> translation within a day, but a human translation would take a week
> or more. Very often, they would look at the article and realize
> that it wasn't what they wanted. If they had asked a human to spend
> time working on the translation, they would feel guilty about throwing
> it away. But they had no guilt feelings about the machine translation.
>
> As for quality, it wasn't a problem because the most important info
> was in the equations, and all the readers needed was enough background
> to explain how the equations were being used. Sometimes, the computer
> made mistakes, such as generating the translation "nuclear waterfall"
> instead of "nuclear cascade". But they knew perfectly well what was
> meant, and they even joked about nuclear waterfalls. Furthermore,
> the computer was always consistent in its choice of words so that
> they knew that the same English words always went with the same
> Russian words. And some of them knew enough Russian so that they
> could check the translations -- they used the system as a convenient
> dictionary.
>
> Re quality: When SYSTRAN was converted from the original IBM 7094
> to System/360, the output was switched from upper case to mixed case.
> Even though the output was identical to what had been generated by
> the 7094, the users thought that the quality had improved.
>
> PC> Do you know of any programs publicly available that actually can
> > do the kind of spatial reasoning that people can do? Do you think
> > that virtual reality and other simulations actually reason spatially
> > in some way similar to the way people do?
>
> Both computers and humans can transform images by rotations, reversals,
> combinations, etc. But there are still many unsolved questions about
> how the human brain does anything.
>
> In any case, the algorithms we use for computing semantic distance
> measures on graphs (conceptual graphs, by the way, but that really
> isn't much of a restriction, since CGs can represent any graph) can
> be adapted to compute semantic distance measures on geometrical
> representations.
>
> Whether those operations are similar to what goes on inside the brain
> is unlikely, but the crucial test is whether the measures that the
> computer would determine as close or distant have a high correlation
> with what people would consider close or distant.
>
> If you can hold your nose to look at Google, you can type "spatial
> reasoning" to get 329,000 things to look at -- or not, as you prefer.
>
> Joh
>
>
> _________________________________________________________________
> 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
> (07)
_________________________________________________________________
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)
|