Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 043
AIList Digest Tuesday, 4 Mar 1986 Volume 4 : Issue 43
Today's Topics:
Queries - Lexicons & Q&A & MAC-SCHEME &
Distributed Problem Solving for Architectural Design &
Chinese Language Environment on Symbolics,
AI Tools - Lisp for 68k Unix World & Rete Algorithm,n
Linguistics - Ambiguous Sentences
----------------------------------------------------------------------
Date: 1986 Mar 3 08:28 EST
From: Bob Weber <WEBER3%HARVARDA.BITNET@WISCVM.WISC.EDU>
Subject: QUERY RE PUBLIC AND PRIVATE LEXICONS
I am currently evaluating available lexicons as part of a project
to develop a NLP system with commercial potential.
I would like information concerning machine-readable lexicons
and thesauri that are not now commercial products but that are publically
or privately available. Specifically, I am interested in the
following information: (1) number of words and how they were
selected for inclusion in the lexicon, (2) how much and what
kind of syntactical information is incorporated, (3) for verbs, whether
case information is included, and if so, what kind and to what extent,
(4) whether the lexicon incorporates any class hierarchy information,
(5) references to research using the lexicon, (6) the willingness
of the owner to share or sell, and approximate price if for sale,
(7) other descriptive information necessary for evaluating the
contents of the lexicon.
Please reply directly to: Weber3%Harvarda.BITNET@WISCVM.WISC.EDU
Thanks in advance. If the replies are sufficiently interesting,
I will repost.
------------------------------
Date: 27 Feb 86 20:21:17 GMT
From: decvax!genrad!panda!talcott!harvard!seismo!rochester!kodak!bayers
@ucbvax.berkeley.edu (mitch bayersdorfer)
Subject: Query: Q & A by Symantecs
On the IEEE telecast on February 26, 1986, there was mention of
a natural language driven database program called Q & A. Does
anyone know of the source of this package?
- Mitch Bayersdorfer
Applied Technology Organization
Artificial Intelligence Laboratory
Floor 4, Bldg 23, Kodak Park
Rochester, NY 14650
(716) 477-1972
UUCP: rochester!kodak!bayers
------------------------------
Date: Sun, 2 Mar 86 19:55:34 pst
From: Harvey Abramson <abramson%ubc.csnet@CSNET-RELAY.ARPA>
Subject: information on MAC-SCHEME
Does anyone have information as to the existence and availability of an
implementation of Scheme to run on the Macintosh?
------------------------------
Date: 3 Mar 1986 22:15-PST
From: hinke@usc-cse.usc.edu
Subject: distributed problem solving query -- architecture
I am currently researching the application of distributed problem
solving techniques to the solution of architecture (houses and
buildings) design problems. I am especially interested in any work in
which multiple agents, possessing different design perspectives, have
been applied to a design problem, While the domain is architecture, the
intent of the research is to investigate the computer science issues
inherent in multiple problem solver design approaches. Reply can be
sent to hinke@usc-cs.
Tom Hinke
------------------------------
Date: 1 Mar 1986 1801-EST (Saturday)
From: Andy Chun <hon%brandeis.csnet@CSNET-RELAY.ARPA>
Subject: Chinese Language Environment on Symbolics
We are currently developing a Chinese language environment on Symbolics Lisp
machines. This includes a basic character set of about 7,000 characters and
a user-interface for standard Chinese character code and pinyin input. This
environment will be used for Chinese natural language understanding research
and Chinese text-processing.
To avoid duplicating efforts, we would like to know if anyone has already
developed such an environment on a Symbolics machine. We are also
interested in knowing other research groups who may be interested in using
such an environment.
US mail:
Hon Wai Chun
Computer Science Department
Brandeis University
Ford Hall 232A
Waltham, MA 02254
------------------------------
Date: Thu, 27 Feb 86 19:48:00 pst
From: bellcore!decvax!decwrl!pyramid!hplabs!oblio!paf@ucbvax.berkeley.edu
(Paul Fronberg)
Subject: Re: seeking lisp for 68k unix world
You might try SCHEME from the GNU distribution tape. I brought it up on a
5.2 box (68020) by a minor modification of the makefile. Also the price is
right considering that this includes source code ($150).
------------------------------
Date: Mon, 3 Mar 86 10:07:13 PST
From: dual!hplabs!tektronix!tekchips!chanl@ucbvax.berkeley.edu (Chan Lee)
Subject: Re: Query -- Rete Algorithm
The Rete algorithm is described in detail on the article(by C. Forgy)
"Rete: A Fast Algorithm for the Many Pattern/Many Object Pattern Match
Problem", Artificial Intelligence, Vol 19, Num 1, Sep 1982.
You can find a lot of relevant papers in the reference of this paper.
Among them, McDermott, Newell and Moore's paper on the "Efficiency of certain
production system implementation" seems very helpful.
chan lee
------------------------------
Date: Thu 20 Feb 86 09:18:01-PST
From: FIRSCHEIN@SRI-AI.ARPA
Subject: ambiguous sentences
Here is the file of ambiguous sentences.
If you want to post any or all of it, be my guest.
From: BATES@G.BBN.COM
The all-time classic is "Time flies like an arrow", which has at least
5 ambiguous interpretations if you allow it to be the first part of
an unfinished sentence (which is how a parser would have to consider it) as
well as a complete sentence. The interps are:
1. The cliche we all understand the sentence to mean.
2. An imperative, as in "Take this stopwatch and time these flies the same way
you would time an arrow in flight."
3. An imperative, as in "Take this stopwatch and time these flies the same way
an arrow would time the flies if an arrow could use a stopwatch"
4. "Time flies (which are like Horse Flies or Bluebottle Flies) are fond
of an arrow"
5. "Time flies (as above), in a manner similar to an arrow, ..." (The end
of the sentence could be something like "move through the air rapidly")
There may even be another interp in there somewhere, but that's what I
remember for now. If you get other sentences that are that heavily
ambiguous, I would very much appreciate seeing a list of them.
Thanks,
Lyn Bates
BATES@BBNG.ARPA
From: Shrager.pa@Xerox.COM
Subject: multiple ambiguity
John made Jim die by swallowing his tongue.
E.g., John forced Jim's tongue down Jim's throat.
John ate Jim's tongue (the rudest version).
John swallowed his own tongue and Jim died laughing.
John ate the cow's tongue that Jim had tainted with hot peppers
so Jim died laughing. It was on John's plate.
<Same>, but it was on Jim's plate.
The tongue belongs to some third person (referent of "his").
From: FRAMPTON%northeastern.csnet@CSNET-RELAY.ARPA
The following is only four ways ambiguous, but the ambiguity is purely
syntatic and the sentence isn't overly contrived. It is a good test of
a syntatic parser.
"I sent the man who is too stubborn to talk to Jack."
The four readings can be deduced from:
(1) I sent X to Jack
(2) I sent X
(3) I sent X to talk to Jack
(4) I sent X Jack (dative shift)
Please either post the results of your inquiry on the AILIST or csnet-mail
the results to me. I'm quite curious.
From: Stephen G. Rowley <SGR@SCRC-STONY-BROOK.ARPA>
One classic example is the phrase "pretty little girls school". One
source of ambiguity is "pretty", which could mean either "beautiful" or
"moderately". However, most of the ambiguity comes from binding powers,
i.e., where you attach the adjectives. J. C. Brown, in his work on
Loglan, gives 17 meanings. Here they are, always interpreting "pretty"
as "beautiful".
P = pretty; L = little; G = Girls; S = school. The problem is how to
insert parentheses into P L G S. (Actually, it's more complex than
that, since you can put in a connective between adjectives to
effectively make a compound sentence; see [5ff]. Also, the some
adjectives can be present in both components of the compound; see
[9ff].)
Binding Meaning
======= =======
[1] (((P L) G) S) A school for girls who are small; the
smallness of the girls is beautiful. [This
is purely left-associative.]
[2] (P ((L G) S)) A school for girls who are small; the
speaker's opinion is that such schools are
beautiful. [Cf. [15].]
[3] ((P L) (G S)) A school for girls; the school is small and
the smallness is beautiful.
[4] (P (L (G S))) A school for girls; the school is small;
the speaker's opinion is that such schools
are beautiful. [This is purely
right-associative.]
[5] ((P G) S) & ((L G) S) A school for girls who are both beautiful
and small. [Both components left-associate.
G is duplicated.]
[6] (P (G S)) & ((L G) S) A school for girls; the school is pretty;
the girls are small. [First component
right-associates, second component
left-associates. G is duplicated.]
[7] ((P G) S) & (L (G S)) A school for pretty girls; the school is
also small. [First componentleft-associates,
second component right-associates. G is
duplicated.]
[8] (P (G S)) & (L (G S)) A school for girls; the school is both
pretty and small. [Both components
right-associate. G is duplicated.]
[9] ((P L) S) & ((P G) S) A beautifully small school for beautiful
girls. [Note duplication of P; both
components left-associate.]
[10] (P (L S)) & ((P G) S) A small school which is thought to be
pretty; also it's for pretty girls. [P
duplicated; association is right/left.]
[11] ((P L) S) & (P (G S)) A school which is small and whose smallness
the speaker considers beautiful; also a
school for girls which is itself pretty. [P
duplicated; association is left/right.]
[12] (P (L S)) & (P (G S)) A small school which is pretty; also a
school for girls which is pretty. [P
duplicated; both components
right-associate.]
[13] ((P L) S) & (G S) A school which is small and the speaker
considers that smallness to be beautiful;
also it's a school for girls.
[14] (P (L S)) & (G S) A small school which is beautiful and which
is a school for girls.
[15] (P S) & ((L G) S) A beautiful school which is for small girls.
[Unlike [2], the beauty of the school is
independent of L & G.]
[16] (P S) & (L (G S)) A pretty school which is for girls and small
as girls schools go.
[17] (P S) & (L S) & (G S) A school which enjoys all 3 properties of
being beautiful, small, and for girls.
[There's another set of 4 sentences that Brown didn't exhibit in his
book. They're of the same class as [5-8] and [9-12], but duplicate L
instead of P or G:
[18] ((P L) S) & ((L G) S)
[19] (P (L S)) & ((L G) S)
[20] ((P L) S) & (L (G S))
[21] (P (L S)) & (L (G S))
That brings the total to 21. However, since we're both getting bored
with this by now, and you've undoubtedly gotten the point, we won't
analyze them!]
One of Brown's points in Loglan was that, in order to be unambiguous,
the language needs pronounceable parentheses and connectives so that the
groupings above become apparent. Each of the 17 (or 21) above meanings
has a separate pronounciation in Loglan; you're not allowed to be vague
about binding of adjectives. (The default is left-associativity.)
One might object that I've left out cues to understanding, such as
punctuation (commas and apostrophes) and tone of voice. That's true;
many cues to understanding sentences like these come from lexical or
prosodic factors like that. However, tone of voice gets lost in writing
and punctuation is lost in speaking (at least partially; consider
"girls" vs "girl's"). Therefore, coping without some of these cues is
still a valid problem.
From: mab@aids-unix (Mike Brzustowicz)
My favorite is "The technician made the robot fast."
-Mike Brzustowicz
<mab@aids-unix>
From: William Dowling <Dowling%upenn.csnet@CSNET-RELAY.ARPA>
Re the recently posted question seeking multiply ambiguous
sentences: the easiest way to make multiply ambiguous sentences
or phrases is to exploit the tree inequality X(YZ) <> (XY)Z.
For example "a book and a stapler or some tape" is doubly
ambiguous, and "a book and a stapler or some tape and a newspaper"
is 5-ways ambiguous. The same trick makes "the man with a hat
and a monkey in pajamas" heavily ambiguous. Of course if n1 and
n2 are noun phrases k1- and k2-ways ambiguous then "<n1> is no <n2>"
is a sentence that is k1.k2-ways ambiguous. Bob Wall once told
me that an early automatic translation program picked up many of
the readings of "Applicants who apply for licenses wearing shorts
From: Walter Hamscher <hamscher@MIT-HTVAX.ARPA>
There's always the old standby "I saw the man on the hill with the
telescope." This is used in Winston's textbook. I count six meanings.
From: John DeCarlo <M14051%mwvm@mitre.arpa>
My favorite is:
"Mary had a little lamb."
It supposedly has at least a dozen meanings, most of which I can't think
of off the top of my head, but I know it is in at least one of my textbooks.
Mary owned some meat from a young sheep
ate an actual live animal
had intercourse with
was accompanied by
...
John DeCarlo
<M14051%mwvm@mitre.arpa>
------------------------------
End of AIList Digest
********************
