Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 111
AIList Digest Thursday, 7 May 1987 Volume 5 : Issue 111
Today's Topics:
Queries - Pattern Recognition Application & IJCAI Information &
Multiexpert Knowledge Systems,
Review - Books on Common Lisp and Prolog,
Linguistics - Style Checkers,
Speech Understanding - Difficult Speech Examples
----------------------------------------------------------------------
Date: Wed 6 May 87 11:22:00-PDT
From: Ken Laws <LAWS@STRIPE.SRI.COM>
Subject: Wanted: Pattern Recognition Application
I'm trying to work up a proposal for research in classificatory
neural networks. The computational end is easy, but I need an
application that will provide the data for experimentation. I
need either real-world data or (perhaps even better) a way to
synthesize interesting data. I'm already aware of some work in
image analysis, speech recognition, and character recognition,
but would be interested to hear from people who have current
problems that are not being adequately addressed. Can you suggest
any other pattern recognition problem or "signal" of particular
interest to the Air Force, Navy, Army, NASA, NSF, or other funding
agency?
-- Ken Laws
LAWS@STRIPE.SRI.COM
(415) 859-6467
------------------------------
Date: 5 May 87 20:35:45 GMT
From: hao!gatech!akgua!emory!cmb@ames.arpa (Chang Bang)
Subject: help me
(1) I am getting no information about IJCAI
in Italy. Help me.
(2) I would like to know an economical way to
attend IJCAI in Italy. Help me.
------------------------------
Date: Thu, 7 May 87 09:59:03 edt
From: dg1v#@andrew.cmu.edu (David Greene)
Subject: Multiexpert Knowledge Systems
Does anyone have information on "multiexpert knowledge systems" (MKS).
Specifically a recent blurb in Business Week (May 11, pg.141) mentioned a
system by Major Stephen R. LeClair of the AI group at the Materials Lab. at
Wright-Patterson A.F.B. The system combines expert knowledge from multiple
domains to solve complex problems in manufacturing.
I'm currently starting work on a learning program which attempts to
coordinate disperate knowledge bases to solve a problem so I would greatly
appreciate any information on MKS (or relevant areas).
Thanks.
dg1v@andrew.cmu.edu
------------------------------
Date: Wed 6 May 87 20:03:29-EDT
From: John C. Akbari <AKBARI@cs.columbia.edu>
Subject: books on common lisp & prolog
>Can anyone make a comparison between Wilensky's "CommonLispCraft" and Tatar's
>"A Programmer's Guide to Common Lisp"? What are the strengths and weaknesses
>of each book? I know about Steele's book but it is in a different
>class.
>...
> Bill Roberts
Having spent some time working with several people in learning to
become programmers, I have a few comments regarding books available
(titles are approximate, don't have them in front of me):
in general, experience points to several important needs when teaching
& selecting stuff:
- students learn well by studying *working* examples, both in terms of
how to program as well as details like style, data abstraction, etc.
providing well-documented examples motivates all sorts of queries
regarding syntax, efficiency, portability, etc. as well.
- the trade-off between presenting concepts & gory details is a
personal matter, but i've found that some amount of both needs to done
at the beginning. once the student begins to think in the
mind-expanding frame of lisp (rather than, say, c), more & more
details can be presented, relying on the student himself to find out
about implementation details on his own.
COMMON LISP
not familiar with tartar's book. the following three are the best I
know of for learning. would recommend using winston & horn and
wilensky almost concurrently initially (personal choice as to which
offers a clearer intro to lisp concepts), followed by the second half of
winston & horn for applications. those interested in learning much
more about hacking should spend more time with wilensky. steele tends
to be useful after one has learned a fair amount about lisp hacking in
general. would use with abelson & sussman _structure & interpretation
of computer programs_ (mit press) for great intro to concepts
(streams, data-driven programming, ...).
steele. _common lisp: the language_ digital press, 1984.
besides being the de facto standard in most ways, it is also helpful
when trying to port across machines (much easier than lugging around
documentation for symbolics, ti, lucid, vax lisp...). also good for
those late-night hacking sessions (steele manages to imbue
mind-boggling minutiae [spelling?] with a dash of humor). not
necessarily recommended bedtime reading, however (except for
masochists).
wilensky. _commonlisp craft_ norton.
an excellent, readable intro to common lisp hacking, with just the
right blend of tutorial & documentation. tends to go much more into
the details of common lisp than does winston & horn, but still in a
readable, useful format. last chapter or so presents a pattern matcher.
winston & horn. _lisp_ (2nd edition) addison-wesley.
good, readable intro to lisp in general, with common lisp almost an
incidental choice of dialect. tends to rapidly leave low-level
details in preference for looking at ai applications of lisp. very
helpful in introducing people to concepts that have been robustly
developed elsewhere (e.g., there is a simple frame system, simple
intro to object oriented programming via FLAVORS, mathematical
examples, pattern matcher, expert system inference engine, etc.).
PROLOG
prolog is as different from lisp as lisp is from c, at least in terms
of teaching it. would begin with bratko, followed by sterling &
shapiro after student has done a bit of prolog programming. introduce
clocksin & mellish somewhere in between.
bratko. _introduction to prolog programming for artificial
intelligence_. addison-wesley.
great intro to the language. presents just enought about
backtracking, unification, etc. along with examples to be of real
value. gives many examples & exercises. not too much humor, though.
examples of searching, expert system shells, games, etc.
sterling & shapiro. _the art of prolog_. mit press.
excellent book on advanced topics in prolog: garbage collection,
efficiency, top-down vs. bottom-up construction of data structures,
etc. many such topics have not been covered elsewhere, and are done
very well. would not recommend to inexperienced, motivated people;
first part tends to be a mathematical intro which may not be appealing
to some.
clocksin & mellish. _prolog_ springer-verlag.
standard text. a little hard to read at times, tends to bog down in
gory details before reader has feel for language as whole.
let me know if this helps.
john c akbari
ARPANET & Internet akbari@CS.COLUMBIA.EDU
BITnet akbari%CS.COLUMBIA.EDU@WISCVM.WISC.EDU
uucp & usenet ...!seismo!columbia!cs!akbari
DECnet akbari@cs
PaperNet 380 riverside drive, no. 7d
new york, new york 10025 usa
SoundNet 212.662.2476
------------------------------
Date: 6 May 87 12:46:57 GMT
From: gilbert@aimmi.UUCP (Gilbert Cockton)
Reply-to: gilbert@hci.hw.ac.uk (Gilbert Cockton)
Subject: Style checkers
In article <8704250321.AA15773@uhmanoa.ICS.HAWAII.EDU> todd@humu.UUCP
(The Perplexed Wiz) writes:
>In article <12295086246.19.HAYES@SPAR-20.ARPA> HAYES@SPAR-20.ARPA writes:
>>Let me briefly add a seconding voice to Linda Means comments on the horrible
>>output of the style-criticising programs illustrated a while ago. That
>>people should suggest using such things to influence children almost makes
>>me agree with Weizenbaum.
>I think that we have two extreme views here. I agree that the style
>checkers available for microcomputers are not very sophisticated. I also
>agree that such tools should not be used exclusively to teach children
>(or any other age group for that matter). However, to say that these
>microcomputer based style checkers have no place in teaching children
>to write in not correct.
A few simple grammatical rules (concord, apostrophes, tense structure,
clausal agreement), as these style checkers stand, you are most incorrect -
and I am even more surprised at such comments when they come from a psychology
grad - unless you're doing AI or rat research that is in which case
you're probably a long way from mainstream psychology:-).
The problem with most checkers is that the rules they embody have
often just been made up by technical writing pundits. As long as they
stick to indoctrinating those engineers and other culturally deprived students
WHO NEED HELP WITH THEIR WRITING (not all do), I don't mind - they probably do
improve the writing of some people from dreadful and unintelligible to
ugly and constipated :-).
However, the minute their jibberish is proposed as something for the
whole school population, then the authority of the armchair
philistines has to be scrutinised carefully. There is not an ounce of
decent psychological research on text comprehension behind most of the
pronouncements of technical writing rednecks. As for literary
aesthetics, this doesn't get a look in - anyone care to stick a novel
through one of these joke programs?
So, the first prerequisite for style checkers in schools is proper
experimental validation of the rule base - breaking/obeying rules
must be shown to have a measurable effect on comprehension
performance.
The second prerequisite is the harder one and takes us into the
Weizenbaum camp - the rules checked in the experiments must be
translated faithfully into a program - not easy as we know that
our current formal representations of language and knowledge are
wholly inadequate, and given the nature of computation may never be
adequate. Philosophical objections apart, I will never trust programmers with
no background in what they are programming to get the job right unless
the domain experts have a cast iron way of validating the program (this works
well for many science and engineering problems, as well as for simple
data processing).
So, the current style rules aren't rules, and even if they were their
encapsulation in a computer program cannot be proven.
--
Gilbert Cockton, Scottish HCI Centre, Ben Line Building, Edinburgh, EH1 1TN
JANET: gilbert@uk.ac.hw.aimmi ARPA: gilbert%aimmi.hw.ac.uk@cs.ucl.ac.uk
UUCP: ..!{backbone}!aimmi.hw.ac.uk!gilbert
------------------------------
Date: 4 May 1987, 16:32:35 EDT
From: Norman Haas <NHAAS@ibm.com>
Subject: Difficult Speech Examples
Two speech recognition trickies from Eng. Lit.:
Our Glass Lake (Hourglass Lake) -- Nabokov
Make-Believe Express (Maple Leaf Express) -- Thurber
------------------------------
Date: 4 May 1987 2252-PDT (Monday)
From: Eugene Miya N. <eugene@ames-pioneer.arpa>
Subject: Updated list of speech examples
For future purposes, I will be placing a copy of my speech examples
lists on the ames-aurora.arpa host. (Don't check yet.) I've posted
them here and for the comp.ai group on usenet. In the future, I
will separate the ACKs as below for possible liability reasons and to
credit the group as a whole. I will update yearly. The last addition
is particularly interesting. See that type of "writing" has a use after
all.
FYI, aurora is an upgrade of a system which originally did speech synthesis
on an old V*x system, so I only think it appropriate it goes there.
Happy hunting with this little bit of `network memory.'
P.S. I was asked for more Japanese examples, so if anyone in Japan is
working on the subject, I would appreciate examples, I won't be going there
until Fall of 1988 (Cray User Group meeting and more). And this appears
to be a critical area.
--eugene miya
NASA Ames Research Center
eugene@ames-aurora.ARPA
"You trust the `reply' command with all those different mailers out there?"
"Send mail, avoid follow-ups. If enough, I'll summarize."
{hplabs,hao,ihnp4,decwrl,allegra,tektronix,menlo70}!ames!aurora!eugene
Acknowledgements:
elman@amos.ling.ucsd.edu (Jeff Elman)
mcguire@aero2.aero.org
minow%thundr.DEC@decwrl.DEC.COM Martin Minow (ex-DECtalk developer)
Marc Majka <ames!seismo!ubc-vision!vision.ubc.cdn!majka>
Joseph_D._Becker.osbunorth@Xerox.COM
Stephen Slade@Yale.Arpa
Keith F. Lynch <KFL%MX.LCS.MIT.EDU@MC.LCS.MIT.EDU>
George Swetnam m06242%mwvm@mitre.ARPA
Erik A. Devereux <GV.DEVEREUX@A20.CC.UTEXAS.EDU>
"In mud eels are, in tar none are".
grey day / grade A
euthanasia / youth in Asia
"Whats that up in the road" ahead / a head?
"Take off your hat and dloves"
and then ask them what you said. 99% of all people will insist that
you said the word "gloves".
I'd be happy if you could do the digits, including "Oh", and Yes/No.
Continuous digits, telephone quality, no training, male and female voice.
The problem is in distinguishing "oh" from "no".
Getting the alphabet (not "alpha", "bravo", but "aye", "bee") would
be nice, too.
I love you
Isle of View
I think you need at least one example in Chinese, and here's my favorite
(because I actually said it by mistake). The numbers after the words
are phonic "tones". What I meant to say was:
Wo(3) hen(3) xiang(3) shui(4)-jiao(4) -- I want to go to sleep
... but what I actually ended up saying was:
Wo(3) hen(3) xiang(4) shui(3)-jiao(3) -- I am like a boiled ravioli
"ice cream"/"I scream"
"beginning"/"big inning"
"soccer"/"sock her"
"its hardware problems are intermittent"/"it's hard where problems ..."
"attacks"/"a tax"
from Mark Twain:
"Good-bye God, I'm going to Missouri."/"Good, by God, I'm going to Missouri."
A notion of water/an ocean of water.
[New York accent only] An arm and a leg/a nominal egg.
Years ago at Bell Labs, I heard the following:
"Joe took mother's shoe bench out; she was waiting at my lawn."
With regard to difficult speech recognition problems, I just saw
variations of the following on the wall of a mens room, so credit goes
to anonymous students at the University of Texas:
``Our understanding of urine formation was clearly wrong.''
``Our understanding of your information was clearly wrong.''
------------------------------
Date: Wed, 6 May 87 09:58:39 pdt
From: Eugene Miya N. <eugene@ames-pioneer.arpa>
Subject: Speech.examples installed on ames-aurora.arpa
For some reason, our news system is not receiving the AI list.
I have installed the file pub/speech.examples for anonymous login.
Someone else noted the lack of AI and cognitive "biggies" donating
examples. Don Norman contributed an additional example from his book,
so if you are concern with this topic you can try ftp. Someone
can let me know if it works. See you all in a year when I ask for updates.
[January]
>From the Rock of Ages Home for Retired Hackers:
--eugene miya
NASA Ames Research Center
eugene@ames-aurora.ARPA
"You trust the `reply' command with all those different mailers out there?"
"Send mail, avoid follow-ups. If enough, I'll summarize."
{hplabs,hao,ihnp4,decwrl,allegra,tektronix,menlo70}!ames!aurora!eugene
------------------------------
End of AIList Digest
********************
