Copy Link
Add to Bookmark
Report
AIList Digest Volume 6 Issue 038
AIList Digest Sunday, 21 Feb 1988 Volume 6 : Issue 38
Today's Topics:
Application - Poetry Analysis & Cryptology and Neural Networks &
Data Smoothing for Character Recognition,
Videoconference - Practical Applications of AI,
Software Engineering - Reference,
Neural Networks - Tools,
Nanotechnology - Semantic Quibbles
----------------------------------------------------------------------
Date: 18 Feb 88 18:10:04 GMT
From: uvaarpa!virginia!boole!cap4r@umd5.umd.edu (Chris Pohlig)
Subject: poetry analysis, pattern recogn.
I have a project that involves determining variations between different
versions of a very long poem. Unfortunately, simple file comparison
programs are inappropriate since not all differences between the versions
are important. For example, many (but not all) spelling variations are
insignificant. Some versions of the poem have extra, or missing lines.
Some corresponding lines (between different versions) are of unequal
length as well. The real need (I think) is to be able to specify (in a
separate "rule" file) a list identifying significant difference rules.
Are there any relevant software products? Are there any relevant
journals? Does anyone have any suggestions?
Please reply to: cap4r@virginia.edu (internet)
or: cap4r@virginia (bitnet)
Many thanks.
------------------------------
Date: 20 Feb 88 00:21:41 GMT
From: farris@marlin.nosc.mil (Russell H. Farris)
Subject: Re: poetry analysis, pattern recogn.
In article <419@boole.acc.virginia.edu>cap4r@boole.acc.virginia.edu
(Chris Pohlig) writes:
>I have a project that involves determining variations between different
>versions of a very long poem. Unfortunately, simple file comparison
>programs are inappropriate since not all differences between the versions
>are important. . . .
Look into using SNOBOL (on a PC) or SPITBOL. A PD version called
Vanilla SNOBOL4 is available--with many sample programs--from
Simtel20. The full-blown version, called SNOBOL4+, is available
for $95 from Catspaw, Inc., P.O. Box 1123, Salida, Colorado
81201, (305) 539-3884.
Russ (just a happy customer) Farris
------------------------------
Date: 18 Feb 88 23:38:24 GMT
From: sdcc6!sdcc13!ln63wgq@sdcsvax.ucsd.edu (Keith Messer)
Subject: Re: Cryptology and Neural Networks?
Yea, Jeff Elman here at UCSD tried to apply neural nets to both the enigma
and DES, and I believe the problem was that the encryption algorithms are
simply too complex. The neural net ends up memorizing the key-cyphertext
pairs you give it, but fails to come up with a good rule for learning new
pairs. I wouldn't discount neural nets for cryptology but they haven't been
useful for straight known-plaintext decryption.
Keith
------------------------------
Date: Thu 18 Feb 88 09:03:50-PST
From: Ken Laws <LAWS@IU.AI.SRI.COM>
Subject: Re: Help Needed With Data Smoothing, Character Recognition
Much of the MIT vision literature deals with data smoothing and
interpolation by fitting mathematical "thin plates" through the
image data. The data I get is usually too smooth already, which
may be why the human vision system introduces the Mach effect.
The question is, once you have smooth data (e.g., if it were given
to you initially) what are you going to do with it? Threshold it?
Detect edges? Segment it? Match it to templates? To generic models?
Take Fourier transforms? Moment invariants? Count concavities
relative to the convex hull?
The vision literature in graphics tends to consider only binary
data, ignoring the gray levels that high-quality scanners pick up.
There are shrink/expand techniques for smoothing and many papers
on how to characterize approximations to straight lines and arcs
on a digital grid.
You should check out the IEEE book list, particularly the pattern
recognition conferences and related books such as "Machine Recognition
of Patterns" and "Computer Text Recognition and Error Correction".
There is a very old book called "Optical Character Recognition" that
still has some good info on recognition by moments and some examples
of just how bad scanned characters can be.
-- Ken
------------------------------
Date: 18 Feb 88 18:53:18 GMT
From: caip.rutgers.edu!pallab@rutgers.edu (Pallab Dutta-Choudhury)
Subject: Re: Help Needed With Data Smoothing, Character Recognition
In article <18128@topaz.rutgers.edu>, clong@topaz.rutgers.edu (Chris Long)
writes:
> I am currently working on a vision project; solving the font-free
> character recognition problem to be exact. I am looking for any
> and all reference sources dealing with data smoothing and
> noise reduction, especially as applied to graphics. [...]
I am working on a project that has used smoothing
as a preprocessing operation for image segmentation. As a
result I can help you find some relevant articles, and provide some
consultation if you desire. The subject has been explored quite
extensively in the literature, for most applications it's just a
question of finding the correct balance between performance and
computational overhead.
Paul Dutta-Choudhury
Rutgers Univ.
e-mail: pallab@caip
------------------------------
Date: Fri 19 Feb 88 20:18:08-PST
From: ELIOT@ECLA.USC.EDU
Subject: Answer to a Question
Ken:
Someone posted a message on AILIST asking about the IEEE
Videoconference, and you appeared to need some
assistance in answering the question. The Videoconference
was not just in NY (as implied by the questioner) but was
broadcast via satellite "around the globe" (so to speak).
The title of the Videoconference was "Practical Applications
of Artificial Intelligence" and ran for three hours. Elaine
Rich provided the program intro and served as technical
consultant. Case studies were given by Bill Blake (DEC),
Robin Steel (Boeing), and Sandy Marcus (NCR). A
specific focus was taken on programs that serve as an engineering
design aid.
Future topics of the Videoconferences include supercomputers,
microprocessors, lasers, superconductive materials, etc.
More infor can be obtained from the IEEE offices, and
tapes of previous presentations can be purchased as well
(the Feb. 18 show will be ready in about a month).
Hope this helps (but don't sign me as Dear Abbey).
Lance Eliot
IEEE Expert, Associate Editor
and
University of Southern California, Faculty
------------------------------
Date: Thu, 18 Feb 88 11:02:28 EST
From: hendler@dormouse.cs.umd.edu (Jim Hendler)
Subject: Re: AIList V6 #34 - AI in Management, Software Engineering,
Interviewing
In re: Software Engineering and AI
Someone was looking for a more recent Jacob/Froscher paper. One such, entitled
`Facilitating Change in Expert Systems' appears in the recent book
``Expert Systems: the User Interface'' which I edited for Ablex Publishing.
Jim Hendler
umcp
Hendler@dormouse.cs.umd.edu
------------------------------
Date: Sat, 20 Feb 88 13:07:06 EST
From: ST401843%BROWNVM.BITNET@MITVMA.MIT.EDU
Subject: Neural Network Tools
Here is the responses I got to mu inquiry about Neural Network computational
tools. Not many answers, but thanks to those who did answer. I deleted the
parts of the messages that included no NNT information. Conclusion: right
now NN software costs BIG BUCKS. I think I will wait for a while.
Thanasis Kehagias
(Follow messages)
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
You can contact Neuralware Inc. at (412) 741-7699 for a commercail version. I
do not know their current University pricing. Also available an excellent
bibliography. -- Sandeep
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
We (at Beckman Instruments) have purchased the NEURALWORKS PROFESSIONAL
software package from Neural Ware Inc. It runs on the IBM PC & clones.
It's a pretty good package, but the user interface is a bit rough and the
documentation is just a little weak. It costs $495 (last I heard), and comes
with a good overview of Neural Computing and several Neural computing
techniques are supported. Their address/phone is:
Neural Ware Inc.
103 Buckskin Court
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
I've seen something called Mac Brain, but I don't know who makes it. I
venture to guess its about 200-500 dollars. I know it works on the
Mac-II in colour and probably works on lower-end Macs too.
Sewickley, PA 15143
(412) 741-5959
------------------------------
Date: Thu, 18 Feb 88 12:23:55 PST
From: uazchem!dolata@arizona.edu (Dolata)
Subject: Nanotechnology - close, but no cigar (yet).
Minsky takes me to task in his last message;
m> Dolata's reamrks about nanoscopic chemistry missed the point, so far
m> as I can see, in arguing that because it is a scanning microscope it
m> is not involved with individual molecules but is more like regular
m> volume chemistry.
Well, yes I did say that;
d> Note that the process involves SCANNING of whole areas, and not
d> individual pinning. This is nothing new, it can be done by standard electro
d> chemical techniques.
That statement is strictly true. Minsky now proceeds to disagree with me;
m> However, the molecular rearrangement was not
m> accomplished by a conventional bulk effect. Instead, it was
m> accomplished by a sub-microsecond pulse applied during the scan so
m> that it occurred while the needle was over a particular molecule.
Carefull reading of the above statement shows that I didn't claim that the
modification was done by scanning, only the PINNING. If you read my next
sentance, you see that I implicitly allow thier claim to altering
individual molecules by pulsing during the scan (the key word is "then");
d> The ability to then alter individual molecules is
d> not very exciting either, people have been doing that chemically with
d> polymer bound systems for a long time.
Minsky takes exception to my whole message;
m> But I see no reason to denigrate
m> the technique because it uses scanning. Simply think of scanning as
m> examining, and possibly modifying, large numbers of points in
m> sequence. What could be better?
I now take exception to his taking exception, because I already stated
the same thing!
d> The exciting possibility was not
d> strictly addressed; the ability to selectively alter molecules in a
d> spatially regular fashion. I.e., convert one to state 1, convert the
d> next to state 2, etc...
I assume my words "spatially regular fashion" to be the same as Minsky's
"in sequence". We both use the word "possibility" and "possibly",
to discribe the task, and so we agree it hasn't been done yet.
And this possibility was not demonstrated, or even addressed, in FF&A's
letter to Nature. Back to Minsky's message;
m> But I see no reason to denigrate
m> the technique because it uses scanning.
I didn't denigrate the work. My final paragraph had the words;
d> I don't mean to completely pooh-pooh their work. It does indicate
d> an exciting new direction.
If I am denigrating something, I don't use the word "exciting". I did say;
d> However, I caution people from either claiming
d> that they did something that they didn't, or from being swept up in over
d> strong claims.
Foster, Frommer and Arnett themselves make the same caution;
ffa> Our interpretation of this process as the pinning, removal and cleaving
ffa> of the phthalate molecules is still open to question...
I respect FF&A's self criticism and scientific caution. AI sometimes seems
filled with people making over grandiose claims, using fuzzy terms to
disguise that what they are doing is reinventing X's wheel of '65, etc...
I applaud clear thinkers who critically evaluate technology rather than
try to get into newspapers or get tenure on false claims. (for example,
Minsky's (I assume it is the same one?) work on Perceptrons is a good
example of critical thinking).
I don't claim that FF&A made over strong claims. I do caution people in
AI-LIST, or other forums, from being less cautious than the authors (who
should know best).
In summary, I still stand by my claim;
d> What
d> they haven't had much luck doing was using mechanical means to create
d> spatially interesting patterns of altered molecules. What FF&A has done is
d> to point the way, but they missed the biggy.
------------------------------
End of AIList Digest
********************
