nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) (03/01/88)
NL-KR Digest (2/29/88 23:51:29) Volume 4 Number 22
Today's Topics:
Re: Garden-path sentences
Re: Ambiguity in 1st person plural
Are pidgins Indo-European?
German-to-English software
S. Johnson ==> N. Webster gaffe
Neural Network Grammars
Re: what is a grammar (for)
uses of grammar etc.
Submissions: NL-KR@CS.ROCHESTER.EDU
Requests, policy: NL-KR-REQUEST@CS.ROCHESTER.EDU
----------------------------------------------------------------------
Date: Thu, 18 Feb 88 18:41 EST
From: Martin Taylor <mmt@dciem.UUCP>
Subject: Re: Garden-path sentences
I *heard* a fine garden path sentence on the BBC news the other day:
"In the second test of public opinion ...." (referring to the NH primary)
I heard "In the second Test ...", which to one starved for cricket
news demands a completion such as " ... England scored 350 for 3 wickets.."
(note the upper-case T).
The "...of public opinion..." caused a mind-jerking double-take.
Note that both versions almost demand that they be heard in the
context of a News broadcast, but the second is more likely than the
first in an English broadcast, whereas it would be almost impossible
in a US broadcast.
--
Martin Taylor
....uunet!{mnetor|utzoo}!dciem!mmt
mmt@zorac.arpa
Magic is just advanced technology ... so is intelligence. Before computers,
the ability to do arithmetic was proof of intelligence. What proves
intelligence now? Obviously, it is what we can do that computers can't.
------------------------------
Date: Mon, 22 Feb 88 10:08 EST
From: Alex Colvin <mac3n@babbage.acc.virginia.edu>
Subject: Re: Ambiguity in 1st person plural
New Guinea Pidgin (Tok Pisin) has a IE (specifically, English and French)
vocabulary, but that doesn't make it an IE language. The grammar isn't IE,
but resembles its neighbors, Fijian, Tongan, Samoan, Tahitian, etc, all of
which have inclusive/exclusive 1st pl.
All but Tok Pisin have dual pronouns. Tok Pisin often sticks the numerals
tu, tri into the pronouns, "mitupela" we (excl.) two, "yumitupela", we
(incl.) two, "yutripela" (you three),
Other non-IE features using IE words are use of -pela (<fellow) as a
quantintifier suffixed to a numeral preceding a noun ("sikspela man na
meri", six men and women).
Taken from "Languages of Asia and the Pacific", Charles Hamblin. Oddly
enough, it doesn't include Maaori. I guess he assumes you can get by in
English.
"Those French, it's like they have a different word for everything!"
------------------------------
Date: Tue, 23 Feb 88 16:14 EST
From: Robin Jeffries <jeffries%hplrmj%hplabs@hplabs.HP.COM>
Subject: Re: Ambiguity in 1st person plural
>Among all the positive anwers to the original question as to whether
>there were any languages that distinguish first-person-plural inclusive
>from exclusive, no-one has mentioned any Indo-European languages. Are
>there any that make this distinction? All the positive answers have been
>from around the Indian Ocean or N. American, so far as I remember.
This is not a perfect example of this distinction, but I think it at least
meets the flavor of what is wanted. Russian makes the distinction at the
phrase level. 'myi' (excuse me, but I don't recall how it is supposed to
be transliterated) is ambiguously inclusive or exclusive 'we', but tends to
mean exclusive, because 'myi c vami' (literally, 'we with you') is used for
the inclusive we. 'c vami' is enough of a frozen phrase that I think of
these as being two different pronouns that capture exactly the
inclusive/exclusive distinction.
Robin Jeffries
Disclaimer: I am not a linguist, and my Russian is pretty rusty these
days, so take all this with a small grain of salt.
------------------------------
Date: Wed, 24 Feb 88 13:01 EST
From: WHALEN <jsw@whuts.UUCP>
Subject: Re: Ambiguity in 1st person plural
In article <2372@csli.STANFORD.EDU>, goldberg@csli.STANFORD.EDU (Jeffrey Goldberg) writes:
> In article <2644@dciem.UUCP> mmt@dciem.UUCP (Martin Taylor) writes:
>
> >Among all the positive anwers to the original question as to whether
> >there were any languages that distinguish first-person-plural inclusive
> >from exclusive, no-one has mentioned any Indo-European languages....
I was wondering, having just joined the net, whether anyone had mentioned
the origin of nosotros in Spanish as opposed to nos and/or nous and nous
autres in French. I don't know what the original meaning of nosotros was
(? nos altros) but it seems to have implied something different from the
original nos.
Jon Whalen @ AT&T Bell Labs, Whippany NJ
ihnp4!pancho!jsw
pancho!jsw@ulysses.att.com
------------------------------
Date: Thu, 25 Feb 88 06:26 EST
From: Klaus Schubert <mcvax!dlt1!schubert@uunet.UU.NET>
Subject: Are pidgins Indo-European?
In the discussion about ambiguity in 1st person plural Erland Sommarskog
writes:
>> Actually one such a case have already been mentioned, at least if we
>> count Tok Pisin to the Indo-Eurpean languages. Tok Pisin is an English-
>> based pidign from Papua New Guinea. In Tok Pisin we have:
>> "yumi" = "you me" and "mipela" = "me fellow(s)".
>> We yet have to see for an "ancient" language of the IE family with
>> this distinction though.
>>
>> Erland Sommarskog
>> sommar@enea.UUCP
Is there any reason to classify a pidgin language as a member of the Indo-
European family one of its source languages belongs to? Isn't this an
overestimation of the Indo-European contribution to the new language? Isn't
it merely ethnocentrism or laziness when we do not even bother to know what
the source languages of a pidgin are called labelling the language as
"English-based", "French-based" etc.? Or are there valid linguistic reasons?
Klaus Schubert
schubert@dlt1.uucp
------------------------------
Date: Tue, 23 Feb 88 15:24 EST
From: Debbie Morley <morley@ncrcam.Cambridge.NCR.COM>
Subject: German-to-English software
Does anyone know of a Unix or MS-DOS software package that does German-English
translation? I am interested in a word-to-word dictionary (i.e., enter one
German word, receive the English translation), not sentence-to-sentence.
Please post, or send any information directly to me. Any information will be
appreciated, particularly on how I may acquire such a package.
Debbie.Morley@Cambridge.NCR.COM
(614)439-0566
------------------------------
Date: Tue, 23 Feb 88 17:11 EST
From: Bruce E. Nevin <bnevin@cch.bbn.com>
Subject: S. Johnson ==> N. Webster gaffe
I referred to a dictionary informed with the religious and political
opinions of Samuel Johnson. Johnson was opinionated, but it was not
his Dixionary that I had in mind. I meant, of course, Noah Webster.
Bruce Nevin
bn@cch.bbn.com
<usual_disclaimer>
------------------------------
Date: Thu, 25 Feb 88 00:56 EST
From: Ken Laws <LAWS@IU.AI.SRI.COM>
Subject: Neural Network Grammars
In the last issue it was suggested that thoughts existing prior
to verbal expression may be syntactically coded with the aid of a
grammar. It was also suggested that natural language may itself
be a primary form of thought, in which case syntax and semantics
are not independent.
I would like to suggest a third viewpoint, one more closely
related to spreading-activation networks and perhaps other
neural-network representations. Suppose I am "groping for words"
during a conversation. It seems unlikely that I am just having
difficulty bringing forth pre-existing phrases, but I find it
almost as improbable that I have a fully formed thought that
is resisting syntactic coding.
What I find most plausible is that I have a set of activated
concepts that I am trying to linearize in order to produce
a sentence. It's rather like trying to string beads. Several
orderings (e.g., passive vs. active construction) may be
competing, and each implies the activation of additional
syntactic elements. This introduces a reinforcement process
that usually causes one particular ordering to win, but failure
to achieve such dominance can lead to the groping phenomenon.
Similar competitions (including goals as well as concepts) govern
word choice and the details of articulation.
Typing errors offer a particularly good way to study such
competitive/cooperative (or relaxation, or constraint-satisfaction)
processes. Sometimes I misplace a letter by several character
positions. This indicates to me that a neural circuit "knew"
that letter had to be included, so it kept trying until it found
a place for the insertion. The inappropriateness of the final
location was balanced by the increasing urgency felt by this
circuit, with a contribution from another circuit trying to
enforce the right length for the word. Other factors, such
as the ease of making particular letter transitions, may also
be active. The final result is thus a composite (or, loosely,
a vector sum) of all the constructive and censoring forces that
are active during the typing.
This model of cognition has some hope of explaining both slips
of the tongue and our ease of producing and understanding
"ungrammatical" utterances. I doubt, however, that grammars
for such processes are useful, feasible, or even meaningful.
-- Ken Laws
------------------------------
Date: Thu, 25 Feb 88 11:14 EST
From: HESTVIK%BRANDEIS.BITNET@MITVMA.MIT.EDU
Subject: Re: what is a grammar (for)
MT> Date: Sun, 7 Feb 88 15:29 EST
MT> From: Martin Taylor <mmt@dciem.UUCP>
MT> Subject: What is a grammar (for)
MT> . . . a grammar (a) is a theory
MT> about the nature of real language, or (b) [is] a mathematics with its own
MT> axioms and procedures for developing theorems, which by chance might
MT> parallel some things that are observed in real language.
For those working on linguistic theory as a science dealing with an aspect
of human psychology, (a) is what is understood as generative grammar. It
would be absurd to do science in a way that "by chance" might parallel
something in the real world.
One qualification is in place: The theory is not about a "real language", it
is about a "real grammar". "Language" is partly a product (or reflection) of
a grammar. The distiction between the language and the grammar is like the
distinction between data and theory in any science.
MT> . . . view that
MT> a grammar constitutes a set of rules that determine whether a particular
MT> string of words is grammatical or not within a language. If her view is
MT> correct, then what is a grammar FOR? What would it matter whether a string
MT> is grammatical or not?
A grammar which only tells us in a yes/no fashion whether something is a
wellformed expressions (i.e. the case in the [b]-sense of grammar above) would
be of no use in understanding natural language. We want the grammar to tell
us *why* a string is grammatical or ungrammatical, i.e. the grammar should
give a structural description (an analysis) of both well-formed AND
ill-formed expressions.
Arild Hestvik
Linguistics and Cognitive Science Program
Department of Psychology
Brandeis University
Waltham, MA 02254
------------------------------
Date: Thu, 25 Feb 88 12:13 EST
From: John Nerbonne <nerbonne%hpljan@hplabs.HP.COM>
Subject: uses of grammar etc.
In Reply to:
Date: Tue, 16 Feb 88 18:13 EST
From: Bruce E. Nevin <bnevin@cch.bbn.com>
Subject: uses of grammar etc
(in NL-KR Digest 4 (20), 2/23/88)
BN> [...] many formalisms that have developed from the Generative
BN> paradigm but do not give themselves the Generative trademark: GPSG,
BN> HDPSG, unification grammars, categorial grammars, and so on and on.
BN> These formalisms have in common that they overgenerate, that is, they
BN> produce structures that do not occur in natural language and must be
BN> pruned or filtered out.
1. What generates?
A little basic terminology: a grammar formalism specifies the
allowable form of a grammar, its rules, whether categories are primitive
or defined, etc. A collection of rules etc. in a formalism is a grammar,
and it {\it generates} strings of symbols and/or structural
analyses of those strings. The goal of (English) descriptive
generative linguistics is to find a grammar that generates just
the strings and structures we accept as English.
But to speak of a formalism as (over)generating is a category
error. The formalism doesn't generate at all--over-generate or
under-generate or otherwise. It just allows the formulation
of rules and grammars whose job is to generate.
Furthermore, it is trivial to avoid overgeneration within any
formalism; just be very cautious in what you include in the
grammar. It's getting things just right that's hard.
2. A Diatribe on "Filtering"
But the remarks above are misleading beyond their imprecision.
The use of SYNTACTIC FILTERS is controversial in linguistics. A
syntactic filter is a device within a grammer that essentially
allows an otherwise correct generation to abort. Because filters
are suspiciously powerful devices, they ARE NOT USED
(outside of MIT's Government and Binding theory). (A final
qualification about unification grammars might be worthwhile:
unification is a technique for combining feature complexes, and
it certainly isn't incompatible with the use of filters. But
unification-based theories such as LFG don't use filters.)
On the other hand, NO ONE expects syntax to explain every
sort of acceptability variation. Some ill-formedness is
due to semantics ("a four-legged attitude") and pragmatics
("Don't hear me!"), and some researchers refer to this as
a "filtering" of syntax. This is a good idea, and universally
accepted.
3. The charge of overgeneration
Now, no one has produced a grammar that generates English exactly;
they all either overgenerate (generate strings/structures that
aren't English) or they undergenerate (fail to generate some
acceptable English) or both.
It's probably worth mentioning that Chomsky introduced the notion
"generative" in {\it Syntactic Structures} as referring
to any attempt to describe the syntax of natural language in a
way complete enough to define precisely what is in the language.
He seems to have assumed that this must be a generating device,
or that it might as well be (which assumption has since been
discredited, of course, but not too easily).
This slant on "generation" clarifies BN's charge: a grammar
overgenerates just in case its specifications for English are
too inexact. The postulate that noun phrases followed by verb
phrases are English sentences is easily seen to "overgenerate"
since it allows cases of number discord to count as sentences, etc.
Put more polemically: the problem of overgeneration is the same problem
as that of achieving the goals of generative theory, that of an exact
description of language.
4. An Alternative?
BN> The Constructive Grammar of Harris (construction-reduction grammar,
BN> composition-reduction grammar, operator-argument grammar) is an example
BN> of [...] a mathematical theory of precisely those relations and
BN> operations that suffice for language. It does not overgenerate.
I can only take this to mean (a) that Con G doesn't attempt a generative
characterization at all, in which case the news that it doesn't
overgenerate is tautological, and not tidings of comfort or joy;
or (b) that it can provide a generative characterization of English
syntax. Since this would be a solution (even though not a unique solution)
to all the problems of descriptive syntax, my confidence in the
scholarly community is such that I think would have heard more of
the details by now. I just don't believe (b). Con G must be a
different game, so to speak.
5. The purpose of Grammar and "Generative Grammar"
A grammar that is generative in Chomsky's sense (a complete
characterization of what is in the language) is of value to, say NL
Understanding systems, since it allows one to arrive more surely at
the genuinely available analyses of NL input. It is a partial
solution to the problem of multiple parses in NLU. The problem
doesn't thereby go away, but is reduced. Harris's work sounds to
me also useful in this regard, and useful roughly in
proportion to how successfully it meets generative goals.
Generative grammars are less practically useful in other ways:
in characterizing a complex human ability, in making traditional
grammars more precise.
--John Nerbonne
nerbonne@hplabs.hp.com
------------------------------
End of NL-KR Digest
*******************