cs 003. 611. - ed
TRANSCRIPT
:>
,TD 142 973
AUTHOR.TITLE
C \INSTITUTION
SPONS AGENCYEEPORT NOPUB DUECONTRACTbRANT1NOTE
EDRS PRICEDESCRIPTORS
1DENTIFIERS
.DCCUMENT RESUME
) .
-
CS 003. 611. .
Nash-Webber, Bonnie; Reiter, RaymondAnaphor-a-and Logical Form:, On Formal -MeaningRepresentations for,,Natural Language. TechnicalReport'No. 36. .
Bolt, Beranek and Newmanf. Inc., Cambridge, Mass.;Illinois Univ.,, Urhana. tenter for ,the Study ofReading: . ,
National Inst. of Education (DREW), Washington, D.C.,..t
,EBN-3.590. Apr 77
400-76-0116A-164242p,.
MF-$0.83 HC-$2.06 Plus Postage.*Artificial Intellisehce; *Computational Linguistics;-Information Prodeming; Linguistic Competence;*Reading Procsses; *Reading Research; *Semantics;Simulation; -Structural AnalySiS .
*Anaphora; *Center for the Study of Reading(Illinois)'
ABSTRACT _
This paper describes alcomputtiOnal approach tocertain probleMs of anaphora in natural:language and argues in favorof formal meaning-representatl.on languages (MRD'S) for naturallanguage. After presenting arguments in favor of formal meaningoespresentation languages, appropriate MRLs are discussed-. Minimal':requirements include 'proviiitns for. cbnstruct±oas equivalent tolamb0.expressions'nd for the separation of de'scaptional and °
asser-fional information types. A series of examplei,is discussed, inorder to insure that,; (1-) is cases wheie the antecedent aflananaphoric expression is not explicitly present', it caR.bft6n bederived through purely syntactic anipulaticnS cf.an R*ropriatelistructured Miq.and (2) the properties propoledAfor a logical MAL makesuch manipulattcins simple to express and to apply. Given anappropriate represe4tat'ion language,-it is possible to deal 'With ,a
rich- class of 'anaphora. (4.A) !
diq
1
.
I4
_ ,*4t
...0.. .__-
- .
********4***************************************4!********** DpCUments.acquired by ERIC include many informal uhPublihed *
-4
* maeriais ,not available from sithe; sources,- ERIC makes every effo -d *,* to ,obtain the best copy available, 'Nevertheless, items of---7mArginai *
* reprodUcibility are oftet encountered aq'thi.s affects thetrality *
* ofd themicrbfiche-and handcopy reproductions ERIC makes avairable *
,*-iria the-ERIC DocuRent Reproduction Seetice (BUS). EDRS' is not *
* reSpoisibqe for,the_quality of the :original,. document.. Reproductions *supplied by EDRS are the best that can be made from the original.
C**#******************.*********-e*********A**************p*********V
4.,
IL
CENTER FOR THE STUDY OF" READING
. ,
- 'TcchnicalRegort No.-3C
. ANAPHORA .AND LOGICAL FORM:ON FORMAL MEANING REPRESENTATIONS
FOR NATURAL LANGUAGE
Bonnie Nash-Webber*Bolt Beranek and NewmanInc.
Raymond-ReiterUniversity of British Columbia
4kpril 1277
U S DEPARTMENT OFNEALTNCOULATION4 WELFARE 0NATIONAL INSfITLIXE OF
EOUCATION
1.
TwS DOCUMENT mAS BEEN REPRO-OVeD EXACTLY AS RECEIVED FROMTH EPSON OR ORGANIZATION ORIGIN..TING IT POINTS OF VIEW OR OPINIONSSTATED DO NOT NECESSAR11-Y REPRESENT OFFICIAL NATIONAL INSTITUTE OFEDUC4TION POSITION OR POLICY
4.,
\,
BBN Report No. .350University of Illinois :, .
..
at Urbana-Champaign Bolt Beranek=and, N'wman.Inc.1.005 West Nevada Street . SO Moulton StreetUrbana, Ildinoi-s-61801. Cambrillge, Massachb etts 02118
1)
$
IN_
.
.
....This research' was supported b the National Institute of.Education under Contract No. MS-NIE-C-400-76-0116 and by the
N National Research Council of Canada under Grant A27642. The
.'':_____,a.uthors.wish to thank Dr. Alan Mackwarth and Dr. William
,----- Woods fdr reading and,suggesing improvements to earlier-drafts of this paper.
2
Abaphora and Logical Formr
;44.
Table of Contents
/;.ostract_
Introdaction
'II. why Logical. Meaning Representations?O .
Semantics3. CombAation
. C. tlec.,resentatin,
W. On Acprobriate,Logical MRLs
A. Lmbda-expressionsB . 'Separation of Description.al and
Assertional in formation Types
IV, C.xambIes
1. ImpIicit,S.etsr. -
rype_ntecedentsC. Predicate AntecedentsD . "Domkes'!
V. Discussion
VI. Further Problems
References
J
3
4
osae
1
2
(3
5
5
5
7'
?
'8
12
14;_17
.1927
29
30\
3.4
Anaphora and LocAcal Form
Apstract--------
rie argue, on general grounds, .in favor. of
YorTal, mea.n1 2g represeritattons' foi natural
language.. we then adopt,as a "forcing function"'
for the adeguacy of such. a representatioa, the
\?roolem Of identifYingthe possible antecedents of
apnor?C expressions. Tnis suggests certain.
2;ructural propertlec f a ,rWresentati4o.n whion
.facilitate the identification o-f oessiole.
\
J.° antecedents. Given an appropriate represefttation-. ,
. .
.1,?,,hgge, Kith such -nrooertiet, it is t'hpin
t.c) deal with a surprisingly rich class 1-of anaphora:
/ o
- 1 -
4
. ,
-d
to.
Anaphora and Logical Form
I. Introduction
Our objectives this paper. tare twofold: 1
1, to provide a computational approach to certain Problems in
anaphora in natural language;
2. to.argue in favor of formal. meaning representation languages.(MRLs) for natural language.
These to objectives' are pot independent. It apprars that the
solutions to ceytain.probleis in anaphora are best forMulated
with respect to an'appropriatery structured logical MRL,.so that
t'Ine structural. entities out of whicn such an MRE:: is composed
.sulgeSt possible -aptecedent.s:for anaphor resolution... . 4ft. 4... .. . di
4
More ''..specifically, We have set ourselves the following.
. problem: what form should a 'meaning, representation assume in
order to facilitate the identification of possible antecedents of
anaphoric expreisions, and what cocpuatiopal mechanisms does
this task require? Moreover, we have chosen to investigate this
problem of identifying a set ol possible antecedents without1
,.isilvoKing general world knowledge. The eSeparate issue of choosing0
the most Sippropriate antvedent from tHis set C4i1,1', in general,
'require plausib_le easoming based' on such gen'eral world
knowledge. We are also aware of instances where such knowledge-
is required even to propose possible, antecedents. Nevertheles
'in this paper, our concern is to, explore the i'mplfcations of a
purelS', ,syntactic -.approach, as 4%/ell as to ascertain it:1
limitations:. It turns `out that a .surprisirigly rich- class of
'
c.,
.-
..,
.
=1%2. =
4
I4
1. t
Anaphora and tiogicad Farm
anaphora, both pronouns and ellipse,. is amenable to such an.,.
. ...
approach, provided that an appropriately structured lOgical MRL
is used. Me shall .find- that the use of such an MRL leads toI.:
--
'partiCularly simpi 'ules -fdr identifying possible antecedents,..
'and . that the- structure of MRL 'can be exproitei3. .
. , et,
.
`cdmputationally 'to preclude.
cd.rtain inappropriate ones. We shall
also find that this task of identifying possible referents is
'intimately, bound up' with an .ability to ,,form appropriate
descrilptions of them, anditnat these deScriptions are, turn,
intimately related to lqgical form.#
II. Why Logical Meaning Representations? /
Although' there is ,universal reement withiZo the_: AI
community that natural language un erstanding ysteMs must
provide someunderlying meaning repr sentation ontp which suiuface
stringsoare mapped., the nature this representation, remains a,,
Contentious issue. One 'aspert this, debate has to 'do with the,
form that this teoresentat n shout take. There appear to be.P
'two points of view: logical orms [e.g., Sandewall, 1971; Woods,
et al, 1972], and sticuctu'T d tletworkSAe.g.j.Wilks, '1975;.Schank,
1975; Simmons, 1973).
0
ihedistitictiOn .etween these alternatives appears tobe, a
c
signifiCant one S ce logical forms"are clearly formal.languages
within which-Mea rags of surface strirs are represented, whereas.
- .
the latter are -66eled ge-'aphs 4hich somehow,rsp.resent these same
6
J
IJ.
Anaphora and Logical Form
_meaningS. .This, distinction quickly evaporates, however, the
moment one observes that a, ,network is basically' a particular.''
choice of representation (at the implementation level):-.f04.4soMe. A
. ,..'(conCeptual level.) logical 'form. We iriterpret .tf:le work- of'----------
2 ., .
-..,
Schubert [1g76] and Simm.cmls end Brute 419-7-1L as supporting this..,
i
point of .view. "'.
Despite this lack of, any, formal disti.nCtion between networksL
and logicalforms, there is a widespread bias within the AL
community against logical 1RLs for naturguage!, [See for
example, Charniak and WiNcs, 1976] . We suspect that there are ,
-
two' Implicit assomptions underlying this antiformal point of
`1. that tie choice of-a logical, form necessarily implies a'commitment: to a corresponding proof theory as one's solecomputation mechanism;
2. tlitiat 1 gical forms must have their "natural" .rApresentationat the implementation.- e.g.,(x) (Ey . Px,y, & Qx,y . must be represented by theS-expr scion ((X),(. Y-) (AND OP X Y) Q X Y))).
Neither of these assumptions, is justified. we have already,
observed that networks can be best viewed as implementation level
representapions for logical forms, and as we Shall show in
Section IV., t computations that we propose for anaphor.1aphor.,A .
resolution within a. logical MRL are in no way based on any kind
of proof theory. c.
If it -is -the case then that logical forms and networks are.
one.and the same, why . prefer- one; over the othe'rZ: we. favor
logical form on the following grounds:
I,
7
. L ,
..
'A: Semantics -.
-. 1
Anaphora and, Logical, Eorm
(
By vir'tue 'of ts being a formal language, edogical form
inherits a. well defined semantics, namely,, its Tarskiab.
i . .
semantics. !This is.
not the case for netwotk representati6ns......._
preSente4 in 'Vacuo (i.e:, without a translation mechanism mapping
the network to.a logic) : AS Woods [1.971 points out,' networks .
'often fall short of 'this tequirement.
B. Computatdonfkr,
The, high level conceptual representation of meaning provided
by logical forms encourages the formul4On of 'approeriate.
processing alg orithms at an equably - high conceptual level,,
independent, of now these logical forms are represented,at the, .
'Implementation leNieL. This provides for perspicuous descriptions
of algorithms, without' specifying the irrelevant, CONS cell
level ,.-
poanter -chasing details reauieed__by- network
® representations. The_examples,of Section IV, illustrate the .ease
with which such rules can be formulated, as well as their.
conceptual clarity.
Representation;C
Therd are tie issues here: representational perspicuity and
representational adequacy: The first is largely a.subjective
matter. We believe -logic41 fOrms -t be more readable and1
c6mprei:le'nsibl-than their corresponding network form's, especially
J
(
when the usuaC.,
l network, are,consider*ably augmeAed in
orderso correctly represent Topical connectives and quantifiers
Anaphoxa-and Logical ForM
...: 411. .*e. 4
wW. ...and their scopes [Schubert, 197; Hendrix,
)1975f
: 'N s,g,' . .
,1.
'he second issue representational7adegilay is far- more,4.4
"'impptant, largely deaA*hg with tfte ability of a given meaning.4 ..
.representation,, language 'to express the meanings of. surface .
.
strings. A closely related issue is that of represerttational,. , ..
. ..closure. Can one tprl, Diom th'given specification of an MRL,,
wEat.can and what,cannotvb,e expressed, within it? Because any-. .
logical ,MR17, has ,both, a well defined syntax and a well defineda
')semantics, it necessarily exhibits a high' degree (i .closure.
,This is not t he case for network representations presented in.
4vao,uo', preopely because they ha'v rro semantics. 111a'ny - of the
Of
network_based.meaningrepreSent'ations in current natural language.
systems. [Schank, 1975; Wilks, 1975 Norman and Rupelharh, 1975]
suffer froll this defect, a ,fact that makes it extremelyddifficultto W
to assess their Content..
. ,
It is.instructive in copnection with the above discussion to
note that these -fiery same issues were hotly. debater] --v;' t944
data base managemept community during the early 1970's. There
too, the basic choice}aas _between network view of data -
IcapAsyL, 1971 and a logical, (57-cO-called relatiolnal, view.
[Codd, 1970] -Moreover, the arguments. advanced in favor of the
relational view were inmany-.ways isomorphk .to thb e we have
made favoring logical form for meaning, representations. t least'
9
6
1 I I:
,
Anaphora and Lo.gicaf'Form
within th4 data base community, the logical view currently=
pEevails, pdmatily -because' its 'high conceptual level provider$
so- called "data- independence," i.e., one's view of, the data is
independentlof implementation details.
III. Qn'A iptvipriate foLs
previouth section, we argued on faily general grounds, . ..
in favor of toemal meaning representStion languages for 'natural'-. . ,
.. , .
) o .
..language. Of course, not just any logical MRI, will do. At the
very leas)t, 'any .such formal language must` provide for* 4#
quantification and the usual logical connectives, but even under
tOese.-requirements there remains a broad spectrum of possible. -
. logical representations,. There are 4x least ctwo diblen'ione to
this speCtrun corresponding.-- to ,repreSentational1 level and
representational' structure. Nith' -respect to evel,
representations in. current systems range from vert"suffacq"
[Simmons, 1.970]' to very primiti'veased ones [bilks,
1975 ; Schank,'19757'. 4lks, in ComputatiOnal S'emanti'cs .[Charniak
dosand Nilks, 1976: 176] provides a good discussion of these issues:
Tin[ _
thts 'pOper,.. we use as iliu-stratiom an MRL that keeps very-
clothe to the surface syntax and 'lexicon. We do so .because the. .
,
computational task that we heve takem as 8 'forcing function" for/
. an 'adequate , representation, namely, ident -ifyig posqible,
-. r,. .
..
santecedents for anaphor resolution, seems not to require a deep -
4--
... , ..
level and is moreover facil.iated,by a' "surfacy" one, at least.
.
for a broad and i-rrferesting class of phenomena.s,.
..
t.0
Anaphora and .Logical Form
. -
Our focus in discussing logical :VRLs is .on their 4*. ,
whiqh
representational, structure. we have.4,found that the need-"Co- -----.. ,
prov,ide appropriate antecedents for anaOhor. resolution suggests,_
D.
,1
.
cerain'.structueal constNgints on possibleNRLswhich greatly. t
facilitate this proCess. In this connection, we emphasize that, .
.,,,,le , are not here propOsing.a fu,lly developed logical MRL. To do....___.. N
.. ) .
..,.
''.
. ..,-so would require, at the very; least, adequate reprsentatIons for.4 .P.' 7.4'g % , - .,
.. -,
tense, modality, mass terms, events, etc. --issues' whtich we' have ,
,.
'..
\,' sp far completely ignored. The MRL used in this paper is-me'5ely.6 ., 4, ..
.,
-1'' a" vehicle for displayjpg certain formal structural properties'.. .,,.
.
we Dave . fouhd necessary for the identification of. , ,
.
antecedents. Our belief is that any fully articulated 'logical,.., ,. .
ARL will have to 15/Vvide these structural units, if it istd deal
effectively with anaphora. Accordinglv,.one wgVbfing .our.,o
. 9.
proposal is SS a set of design constraiitsorr the structure of
possible logical MRLs.for natural language. - The 'rema inder. of
this section deals with these structural properties.
A. Lambda-expressions-, ./..
' For. ,a formal MRE,.to be adegUairrecor the resolutiom of'vbrb, ,. -. ,
phrase ellipsis, it must' provide tor constructiktsequiValent to.- . -
lambda-expressioits. For dkample, the sentence pair
la. aphn love. Mary.2. So does Qin. .
1
\,10 ,
,. /
, .
requires, as the antecedent., / (),./
-)
the ellipsed verb et the'. /. ,
. /.
/1.formal construct X(x)1Love xf/Mary] corresponding to ,-.gloving
., a .. . -'Mary", whence the resolved sentence -lb. becomest ,. . . ) ...
/-'Anaphora and. Logical. Form .
-. Bill; X(x) ELove x, Mary]
)
which simplifies tot ' ,;Love Bill, Mary
I
,
. .
.
(Note that our preferred notation NN. or, applying //a., ) . .
- lambda-expressibn to an argument is to follow the argument by:the,
.
lambda expression corresponding to normal subject-prediCate word--,.....!-
., , ..-_.. -.....order in English.). .
,
. .,-t -,. "
7/
-Be Separation of Deriptional and Assert.4ional Information:Types ,.,
' Since the "antecedenfs "of, .many". anaphorid pxpressions are,
descriptions, an adequate forrnali.sZ must be. sd, or,ganited at. ,,
e- .,- these descriptions. stand' Clearly, For exapple; confider t4e,\
Pa!I,
- - .
. . . . '.. .
. .
-.pair of s,entence'g. -
. N.,, "-0'
. 1 .
/
. ,>.... ' i. - .. ' '- 5. 't . e-'`-:--,. ',- -,, .-.4. Sqmd. cotton` T-sh-l"rtse.,are expensiVe. ', - .
t/.-
,.. b. but not theOne that Mary. g,a,ag to 'goy yesterday; '-",-4-'\-f< I f,
;,
1n-a "fiat ". pregj.c4ate\Calpulu.s Mlle (ignoring the dietinct,in,
, .,,between "!some" plural and ",Some" singular), sente might be, ,
represented by V
.t. ..
(Ex) . Cotton x 'E, T-shirt k '8,.expeniive x)
. `, k. . l ,ft ;' . ft
/s
I 9
..
I.
'.
I ,
-a
Anaphora and Logical! For
1 ,.
Now intuitively, the ante edent "one" in sentence 2b. is
'. something like 1!ca.t'ton T-shirt", but fro ii the flat predicate
'calculus representation, there is no more reason to suppose, that. - .
4' I.. 't
Cotton and,T-Shirt form a possible antecedipt than Cotton . and
Expensive, cir T-shirt and Expensiveorany one or all'three.,
That ii,' there is no structural indication that Cotton T-Ashirt is'
areferenceable unit. We believe such an indication 'is necessary-
' in any fo''malism adequate for, anaphor resolution..
Usin the structure of a typed logic, predicates that .
constrain the- range of a quantified variable - i.e., types, -7- ---
(like' _1+ -shirt here) can be structurally distinguished from
predicates that aSsert_ things (as -:"Expensive" does here).
Moreover, using the lambda operator, the notion of type can be
eterided trom simple one-place predicates to more complex ones to
yield all and only the allowable referenceable entities.'
For example, we can represent 4.' ^`
rre
O
"T-shirt" i as T-shirt.
"cotton T-shirt" as ),(31T-ahirt)Notton u]. .
.
"T- shirt thatcMary gave, Fred" -as )\(1:T-shirt) [Gave Mary,Frediu]
(The' first s merely a shortie/40 for X(u:T-shirt) (True]'.) Notice
that we are pdatulating .a representation for "cotton T- shirt"'
that is mor,highly structured. than a simple conjunction of
'Cottonand T-shirt (T-shirt x & Cottpn'x). Specifically, we are
separating that part of the noun phrased noting the. primary
1
,r, .
, .
, ,-. 10 -'.
1-3
J Anaphora ana Logical Form"it
. =
.class to which an entity belongs (usually the head noun) from
those.parts-denotiftg restOctions on that )class (conveyed by4
adjectives and relativeClauSeS).
. /b,
.This provii.des yet anoth r structural property that a logical.
,MRL -should po sess ifi on 'to facilitate'the identification of
antecedents forY'ahaphor tkon. Consider
3. Mary bought.a tie -dyed, cotton T-shirt' and Fred bbvght an
embroidered one. ,
whether intuitively "one' refers tcr"cotton T-shirt" or "tie-dyed
cotton" T-shirt" or merely "T-shirt", it must refer at .least to.
"T-shirt",:the primary class denoted by the noun phrase. There.., .
.
i.s no way (pronominally) in English to refer; to a restrdclionI
r
mg
witholit\alsoreferri:ng to the primary claSs. This is Our- main
reasonFfor keeping them distinct in",our logical MRL
CO"ta
4.44. Another consequence of this separation,of descriptional from_
assertional information is that it avoids problems that Woods
1119751'4ecusses with respect to adequate representations for
relative clauses. First, sentences like "A dog that had rabies
bit a man" and "A dag that bit a man had.rabfes" can be assigned
distinct representations, for example,
(Ex : X(u :Dog) [Have u, Rabies]) (Ey:,Man): . Bit x, y
(Ex:X(u:Dog)[(Ey:Man). Bit ,u, y]) . Have x, Rabies'
.
,., More importantly, processing_ rules such as those propoied in
Section IV, can treatat4se two representations differently. As
14
4 ,
Anaphora and LOgical Form
A
woods point but,)conventional semantic networks fail to represent
the distinction between :tilese two sentences. From 'the. -.
p.erspec,tive,of the above discuSsion, .one reason for this. is
clear: a conventional network is ,a repesentation, at the.
.
implementation level, of a "flat" predicate -logic.
IV. Examples
In this section, we present several fragments of ,discourse,.o 4
each containing anaphoric expressions - pronouns and/or ellipses.
. .
Kecall tnat we are not concerned here with the kinds-df exteYnal
knowledge needed to choose among possible. antecedents _fc4 an
anaphoric expreSsion. what we are concerned with is insuring..
that
in cases wnere tne antecedent of an anaphoric expression isnot expliitly present, it can often be derivedthroughpurely :syntactic menipultions of an appropriatelystructured MRL;
2. the properties we have arohosed for a logical MK, mjke suchmanipulations simple to express and apply.
,Since dreveloping our approaCh to anaphora and logical form,
we have discovered that it is compatible with much that is
current intransformational linguistics today. (FOr a survey of
current ideas on anaphora in linguistics and psychology, see
[Nash-webbe, 1977]..) With respect to a level of "lOgical form",
Chomsky [1975] has argued for such a level within a two-stage
system of semantictinterpretation". , In this .system, ,surface
structures 'are first converted to logical forms by semantic
12 -
1.5
-Aniptiora and Logical Form
interpretation rules-involving scope,, bound anaphora, thematic,
'relations, etcr. [Chomsky, 1975: 101]. These logical forms are
:then subject to further interpretation'by other semantic rules
involving, discou rse properties, situation, communicative
intention, [Chomsky, 1975: 104]. to give 'fuller
r
re.presentetions- de- meaning: Moreo,v-e-r,- a primary reason for
poseOlating,such a level seems to be Chomsky s feeling that the
"general principles of anaphora apply;to logical* forms ratherli
than to surfaeestructures directly" [Chomsky, 1975: 241 ft. -31].
Tne 'notion that verb"phrase deletion makes reference to a
logical representa -tion of the sente ce in order)to identify,
"identical" predicates, later instances of which may be ''deleted,
has bedn adva*ed independently by sevekal linguists, including
,[Sad; 1976r Williams, 1977] ploreover, the logical forrgadopted
by these two has a form simiPST to 'Church's lambda 'calculus..20
,Even the nation that pronominal antecedents may not be present;
linguistically, but may have to be derived can be found in the
current linguistics litexatu as well [Bresnan, 19,71] .
With our-examples, we (Aye a small set of manipulatiAon rules
which yield the needed antecedents. We make no claims for 'the .
,
completeness of these rules; there obviously remains a great deal
of woreto be done along ..l lines (see Section VI). 6 Wd do
believe, however, that-the examplesindicate the utility of our4$
basic approach, and that this approach provides a promising
-'hrection for further research.
13 -
t6
f
. e'
'`Anaphdra-and 'Logical Formt
A. Implicit Sets
our first example illustrates One way of'derivip a s@t asa
LAndidate antecedent for "they".4:Consider the sentences
4a. Mary gave each boy a T-shirt.b. She bought t,heir; ae"Macy's.,
. . .. s / ,
The first may be represented aS
4c. (V :Boy)(Ey:T-shirt) . Gave Mary, x, y
tFor simplicity, we will ignore, the fact tna,t reach boy" i$
. ...:c.)ogbi,/.anaphori.c, referring to 'each- .boy in some previouslyS
rontioned set or one implicitly- defined by context,,and treat it
r:tler as a,universally quantified noun phrases)
Notice that we are considering each sentence 'individually,
since we want, to assign it a representation that is correct, but
i
wnioh does not depend on what may follow. , Ihe'result, will often-,t
.. .', ,-
oe a reading that is ia some sense noncommittal: _it wili'b,e' vague.- .,.
:mot true. If subsequently we'learn mace about the : situat4on, we.
./ .---
will. refine this Tepresentation to cefl nect our new knowledge. . -
state, as should become clear, through theqollowing examples.,t.
a. ,'. r ,
The secdn'd sentence- we reprfserit initially with ."its '-,
(,-' anaphoric elements overtly marked, th t is,
4d. Bought SHEl, THEY1, Macy's
(We subscript the pronoun symbols merely to keep several
instances of the same one distinct, aS would be the case in "They
Yi
4
Anaphora and Logical Form.
tilankpd ne tor thee.) (14ext, we identify !possible referents for
tAb anapn
trivially
O ric terms. Since nary is the only female around, we
ssign ner'as tne referent of ShEl.
R63ar ing candidate,anteceJents for il'r""lme-postulate two
1.
,,
, l
ways df.deriving possible sets from sentences like 4a..
;
,iI. Form; tne set description of any ,type restrictini a
uni,rersally auantitied variable. (,Ne reoresent 'the setJesciption of type C.ov (xlCx1.)
, ..
G. Let A oe a (pCior) formula not containing the anaohoricelement '1hr.,Y1 nor any negation in .tie main clause.. (In thecurrent exampte 4c. olays the role of .) Suppose v, has an.1xistentially quantified variable y that, 'lies Within the
. cope of a universalayquahtified-aridne. For: ;1 the set
Jesc1 iotior1 of tne set of y's satisfying'%A. Tnis is a5trai3httorward procedUre, -igvolving the type of y
:restr cted oy an expression'dering from v./. Details atejiveh in k4asn=i,eoper, fdrthcollingl, but the examole shoull.iugye t its oasis outline.
anus, sente ce 4 . yields (x1Boy x), the set of ,boys, Via' the
first prose ure, and (vIT-shirt v & (Ew:Boy) . Gave Mar y,w,v) via
tne second one, i.e. tne set of 1-shires, for each of which,bP
mere is soml cloy to whom Mary gave it. Substituting each of'
thlSe sets turn fdr THEY1, yields
4e. BoughtMary, ixIBoy XI, Ndcy's
t. Bought Mary, fvIT-snirt v & (5.w:3oy) . Gave Mary;w,v), Macy's
That is, either Mary 'bought all the boys at acy's or she ught
there-all the 1-shirts she gave out.(li Real- world knbw/ledoe
would now be needed to dhoose the-moe plausible reading.
(1) -
Representatio n's 4e&f. are omewnat siliplified with: respect to
15 -
18
Anahora.and Logical Form
`Notice that in English the pronoun "they', as well as many
'plural noun phrases, are ambiguous between' a .collective reading
Call together") and a distributive, one ("taken one at a time.).
Sometimes, a lexical item will indicate that a plural should be. e
understood- distributively', as "each" does in "Mary's suitcases
'were each weighed at the airport" Sometimes; semantic
selectional restrictions can be used to choose between, the two.
For xaMple, 'pile requires a collective interpretation of its
object: "She piled them into a heap" doesn't mean "for each one,. -
1
she piled it into a'heap". But often, only unknown aspects of
the situation can furnish the appropriate information.' For
example, if we learn that "Mary's suitcases were weighed' at the
airport"' w6".4cannot say'tor certain whether each one was weighed
there separately or just the whdle lot together.
In line then with the policy described aboVe of always
opting for, a -vague but true interpretation, rather, than making
unsubstantiated choices, we will interpret plurals
'non- distributively, unlers or until additiOnal information would'
lead us, to an alternate decisibn.
Xe
6. I.
substitution of their antecedents), since it 'is- important todistinguish whethee the original representation contained ananaphor or a full term (see Example C.2 below).
-
It
r 0 Anapflora' and Logical Form'
J.: type .Antecedents .
i
1
0(irrl-extexemple,illtistrates-the identificatioh of 1 .
I ;
descriptions as-candidate .antecedents for anapfloric' ",7ne".
1,;.::
4A/5a. Mary-gave each ooy.a green T-shirt..o. She lave Sue a. red one._
4, interpret sentence 5a: like 4a: above, except for thei I
aitional-modifier "green", oh ETsnitt..
A
(Vx:hoy)(Ly:X(u:1-shirt)k3reen u]) . Gave Mar-y, x, y
por-itcn,.:e 5o. can ac represented initially as
(Lz:X(u:P?) ul) . Gave Sue, `zt 1
P., V
i.iat is, there is s=ethin.3 Qf un4,41.,tme --4,3","(tnat should ile
.cP \ , x' t . ,...
erivaole from context, ',which we know exo( ricitly is red, which.,-
some Known female ShC, gave'Sue., Ours task 'cis naw to identify'.s ,'
, . ., ..._,
,
'bossi.ole antecedents Oor ShLi and P?,. -
LI' transformational grammariens, inclUding [tangacker, 1966;
. -> .
heinhart, 1976], present, us with a simple syntactic criterion for,
.rejecting Sue as an antecedent'for Shy "she" preCedes "Sue" in
tne .surface sentence and the node in the parse tree for this
sentence off which "Sne" hangs neither, "command's" (Langacker'S
teFm) rgr "C-commands" (eirihart's term) the node for "she". So
again oy default, there being no other females around, we assign
iary as the referent for SHE1.
17 -
4
4
J..>...7. .,__1___,,,_..__.
.10 - g-4,Anaphora and Logical Porm,.
fi
,
/. . oo
''As 'for. P?, its. possible anteCedents'ihclude all"recently'
mentioned 4 types, independent .of the particular ,guantifiers:.
('- `Recent" seems to mean'here'the 'qurrent sentence, the,prevPous1 .; . ., .
.. , .
. or)e, and- perhaps the one before, t'hat. It does not seem' to . be.
P
a
affected bytask structure [Deutsch', 1975) Or: story structure, or,v°. . _
any of bne other factors' that seem_to change the set of availablepI .
antecedents for definite.konouns, "he", "it", etc.) ,
. h
-' The tyoe's explicitly given in example 5 are': Roy, T-shirt
and X(u:T-shirt) [GrAen-
u). Notice that .when one itype- is,..--
T
constructed Out of other types via the laiabda operator, we
include them4all as posible candidate antecedents. Prescribing
exactly what criteria one would use to identify the most .
plausible antecedent for P?, or in Whaway one Wouid'apply them,
is 'not-within the scope° of this paper. But they would include4o.
th'e semantic criterion that one be Ole to'predicate Red of an, .
,
.
,
..,entity of type PZ `Phis would eliminate Mu:T-shirt)[Geeen y]
through application of a "clashing. color" axiom:, if something is
green, it is not red. (Noti 0.1etra-t----if sentence; -5b. had been
5b.. Fred, she gave anextra-large, one..
,--.... -.there would b-e no"reason to, eliminate this.% description as ,a
* ,. .
Prausib* antecedent.) 'Under rhetorical 'criteria, we would ,)- .
o
expect PaW 121ismito' argue for plausibility. That is, if two
Successive -sentences ,are structurally_ similar ("parallel") and idle%
the latter, anaphoric. "one" )felp-s to fill role. R (here, the
object),- then it has a very '-plausible' antecedent in the noun
V
- 18
.
.1)
,
Anaphora and Logical. Form
phrage. 'filling .role R, in the previous' sentence. (here, the
previous o bject t' green T-shirt") . BUt our point here-is .n.c5,to
specify procedd'res tor/of-loosing among 'candidate anteCedeftts; .it
.
is rather to show how 4_suitable logical fraglwo'rk provides in ,4v.
straightforward way all and'only the approptliate possibilities.
C. Predicate Antecedents
-1. Simple Verb Phrase -Deletion
CP 6
CFO
The 'next few examples ill--1
ustrate some problems involving
verb ...phrase sis,, which are handled rather neatly within .our
framework.
6a. Mary gave,Sue'ab. Jane did too.
The represetation that wee-assign to sentence 6a. is
,(ExLT-shirt) . Gave ,Mary,, Sue, x i.-.
, Sentence 6b., we' interpret as predicating something (P?) of .Jane
previously been predicated of scneone else:.
. P? Jane
.., . .
To identify possible antecedents for P?, we find the.
,
one4lace predidates that either are .giv&n.explicitly or can be,A. \ .tdec v d via lambda absteation on the subje ct position.
... .(Again,.
,../
.
.,
one probably need onliv,seach foe 4h predicatei in the current
CS
sentenceif it has several clauses or i.n -the one ar tWo sentences,. 1,
- .
EK
?-9
.
4;1
,
Anaphora and,LogitcalFcirm
immediately preceding ft, as tIte half-life-, of predicate
antecedents,. !lik that of type antecedents, seems to be very
short. Note that we are viewing the first argument place of a
.preticate as ,corresponding to surface subject position. Though
-this 'reguires a differea representation for active and passiveJ .44 4
. .
sentences, we- see--the need for this on other grounds, for
.example, their difference with 'respect too . siffiple- verb 'phrase
deletion
John hit a linguistrr6a did -too.
as opposed to -
-John was hit by a linguist.Fred was too.
*Fred Ti-J-E66.1.
This example .is' simple in that there is only-one such
one-place predicate-abstractable off a subject r,,
X(r),[(Ex:T. -shitt) Gave r, Sue,
that is, giving Sue Substituting -for P? yields0
Jane,----X(r)[(Ex:T-shiit) . Gave x]
which is equivalent-to6 4T.
. o( t1.0
°. '1/24
to that this representation does not coiimit us to both ,girls4
S 'havihg given Sue the same T- shirt,,- nor need they be .di-. r.
.
The description of the° -first one is,
(Ex:T-shirt.). . Gave Jane, Sue "x
%ri
A
e_
20 -
46.
t
y .. Anaphora and Logical Form46
. (V: T-shirt z & Gave Mary, Sue_
., . e
'.
"a T-shirt that Mary gave Sue",,where r indicates the indefinite
operator.. (z might be called ih°English "the T-shirt which' Mary .
-
gave .Sue" if no other T-shirt in the discourse meets this..Wifr
description".) The second T-shirt ..s desctib.able asos-on
Ylw: T -shirt w & Gave Jane, Sue w
"a T- shirt that.Jane gaVe Sue".
It is important to be. able ,tb .derive- such descripti,bns,
since the entities they- describe may serve as antecedents for-,-0
2,later anaphoric expressions,, for example,
6c1 Neither of them fit her.
. ,. -.,
.
where Ahem" refers to the implicit set of IT- shirts g iven to Sue,*
who is also the mot 'plausible antecedent of'wher".
1,,,
/. "Sloppy Ideritity
Our next example illustrates a phenomenon that has' been
..4,-.,called' the "sloppy identity Problem4 [Ross, 1967]. It- plyolves A. ,
.
acco- unting for . the appearapce of an addiAjorial" .teadin7 for,,
tentences dontaining deleted verb phrases.,, That is, while'
sentence 7a. seems unambigAs, sentence 7b. might mean' either
that Fred beats G wife or thA"56-1)Dafats'hi.s own. How do we,
4 J
account for t sZ,4
.
aP
21 -
4
7a. Garth beats. his wife.b. Fred doe's too.
Anaphora and Logical Form
We assign 7a. an initial representation in which its .
anaphsric term is overtly marked. ,.,
-7c: Beat Garth, 's(Wife) HE,
(wheee- s'is'.defined to be a function that takes a unaryt .
.1. y
predicate lake , Wife, School, etc., and returns a function like., .
uwife..-roi", *"School-of", etc. 'scWife), -for' eXample, is a
-k--, function that takes a mark as its argument and returns his wife:.
, 1,..
,-.141(wife) Jo
.hn is John
,
s- wife. Having a functiOn' like :s, * , ,
,
'-eilminates the need to pbstulate a'sepanate- "Y-of" ,tunction:Orc
.
every unary predicate Y.)
with no'.other male around, we can assign H81 to Garth by
default, .that is,
7d, Beat Garth, '(Wifel Garth
We assign sentence 7b. the representation
6C%114
7e. P? Fred
I
-
O
4
Now, while there are no explicit one-place predicate's around to
serve as an antecedent'_for P?, there are two .ways in which to
abstractone from7d.
(i) >.(r) [Beat r-i--'-slowife) Garthit(ii) >(r) [Beat 's (Witte) r]
F
- 22 -4
ler
Anaphor a arid Logical- Fotil,
)The first repraggents beating Garth's wife and the gecond, beating
one s own. Substitutipg for" P?; we get the two plausible
readings. .
Tre, (r ) [Beat , 's (Wi fe) Garth]Fred, ).:(r) [Beat r, 's (Wife) ri
4
'1'
That is, either Fred' beats Garth" wife \or he beats his own..
i
(We noted earlier the need . to,it distinguish whether an
argument place was originally by an a aphoric expressioh
or by a f:ull noun phrase. Obilerve ;that if sentence 7a. -had E60
"Garth beats Garth wife", which wou164be represented simply as,.
). -.,7d., the following sentence, "Frad does tool', 'Apuld only, mean .
that Fred beats Garth's-wife. Wfille we have simpLified things..
..' / .,
4 . ii,
for, this presentation, in, an *acts l -im6lementation, A would tiave. A 0----
to indicate , ",he = Garth", rather-than simply relp1;eirrg " ,c4.40
, "Garth ":, in order !`to d-erived/ an ''and only the correct 4.amlida
a gtractigns.y
, ,..: ,
, .
. .r ;
k3. Abstractirli -es f 4 O.rom Non PO1
f -r
isI.The point of- the next exarapre s to illustoOd abs racing.
----4,,..5
. tne-:place \'-,ppe-C1i a From positi?ns ,,ot er than the : Urface,
it. ,.,
subjeptv: -JAI %,,h4 `"IikewNe',..:too' "similarly", etc,,,,
..
AI ."----P
A ( .
10
( f 6 1 *We 01- optio a y by epreposition ..anrth-en -by a nce.U6 phr,Ae
) in. digote plat, pie *filled I by. -`the IMW nouri phrase in a :----
.
v 0.previously mentiorreA.-preaicate is a none subjects on 'Uis.
illustrated, ear ier,, a noun ...phr-ase# followed by 'a
Tr,-,6%,,,,e (4.,:... ,<---" !?.j'-,,,_, ,
. ,
\ J i. ,
.- ----- 9 ; 6
,-at.ixiivia,ry, 4.,
Anaphora and Logical Form_s'
requires that the noun phrase fills the subject role of a
previdUrpredicate.)
8d-.---John hit a cop'.b. Likewise, a CIA agent.
The first sentence may be represented as
8c.' (Ex:Cop) . Hit John, x
and the-i.U.econd one as
8d (Ey:CIA-a4pnqi.
where (9? stands for an anaphoric predicate like P?, bilt one whose
IP
0
'argument fills a non-Subject role.
To resolve ?, we must. identify -the one -place predicates
thdt can be abstracted frdm non-:subject positions. FrOm 8c., welf.
get
>40[Hit'John,.r]
'.whibh substituted for O?in 8d., yields
(Ey:diA-agent) . y, )%(r) Hit Johns, r
:that is,, "Likewise, John hit a CIA agent."
It might appear that one could resolve 'likewise"-elli,pses,.
-4- .
at the-. level of the surfaCe.strinT alone, but this is not the
case. Consier the following example:
4,4
Anaphora and Logical Form
9a. John gave Sally her present..b. Likewise, Fred.
ObviQusly, while substituting "Fred" for "Sally" in the surface
string would yield an interpretable sentence, "John gave Fred her
--.present", this is not the only, nor the most plausible reading of
the ellipsis in 9b.
We represent 9a. as
9c. eave. John, Sally, 's(Present)*SHE1
which-we interpret as
9d. Gave John, Sally, 's(Present) Sally .
she being thA only- female around. Sentence 9b., we represent as
9e. e? Fred
To resolve Q ?, we identify the one-place predicates that may be
abstracted froll non-subject positions. From 9d., we get/.
i.X(r)[Gave John, r, 's(Present) Sally)ii.).(C)[Gave John, r, 's(Present) r]
,iii.),(0[Gave John, Sally, r]
substituting for (j)?? and flattening, for clarity, we get
^.
`Gave John, Fred, "s(Prese4t) Sally"Likewise,. John gave Fred Sally's present,."
Gave John, Fred, 's(Presentlo,Fned"Likewise, John gave Fred Fred's present."
Gave John, Sally, Fred"LikeWise, 1ohn gave Sally Fred."
Again, the preferred interpretation would be chosen by using
wotldrhowledge..
Q1.4 C.)
1,
Anaphora and. Logical Form
4. Abstiacting Conjoined Predicates
Our final example of predicate anaphora illustrates another
necessary way of deriving a candidate antecedent: by, first
conjoining predicates, applied to the same argument and then
abstracting a new predicate oft the common argument. That is,
given O
P x1,
- followed by (or explicitly conjoined to)
Q z1,
we can derive
.
y,-X(r) [P xi, & Q zl, r; ...]
4
To illustrate the need for such a rule, considdr the example
10a. I walk an I chew gum.b. Ford does too, but not at the same time.
0
These we represent as
Walk I & Chew-gum I
P? Ford
'(Since we have not introduced a representation for tense, we
cannot represent "but pet at the same time". We,phall Use it
informally, rather, to constrain possible antecedents for P?.
That is, ip? must, sensibly refer to.tWo or mcre actions which are
not done by Ford simultaneously.)
To resolve PT, we must identify the' previous ,one -place
predicates. Walk and Chew-gum are given explicitly, but
26 -
Anaphora and Logical Form.
,
substituting either one for P? leads to an unsatisfactory xesult,
'neither being compatible with,-"but not at the same time" (e.g.,,
"Ford can chew gum, but not at the same time.")%. However, the
above rule yields another one-placepredicate, namely
Mr) [Walk r &-Chew-gum r]
which is a plaugible antecedent for P ?..
D. "Donkeys"
,r.
As our final example of how an appropriate logical
representation of a sentence can yield antecedents necessary for
anaphor resolution, we will consider a', particularly bothersOme
class of-sentences, illustrated b4example 11.
11. Every man who owns 'a dohkey beats it,.
. The pfoblem lies in identifying the* antecedent of "it".: It is
.,
not "a donkey". The sentence does not mean that every .man
owns a donkey beats a donkey, but rather that he beats -any,:donkey/:
that he 'owns Moreover, there is no way of construing the
existential- quantifier associated with "a donkey" sudh that ".it"
falls within its scope.. How does the, correct antecedent for "it"
emerge from our framework?
We f1T-St, -6S-sign sentence 11 the interpretation
(i)' (4x0,01:Man)[(Ey:Donk&Y) . Own o, y]) . --Beat
- 27 -
ft
I
t I , An6PhOra'r'and Logical Form
That 'is, for every man for whom'there is some donkey that he
'owns, he beats it. Now while there ;is nothing explicit to serve
as the antecedent for "it", it turns out that "it" can also
reference a certain kind of functional entity which arises from
t*
We postulate the following rule fortidentifying, a possible
'antecedent for ITI
1. Find a type restriction which -contains an existentiallyquantified variable y, not within the scope of either auniversal quantifier or negation.
2. Determine the description of y with respect to this typerestriction: any entity which satisfies this desciiptioni,sa possible antecedent for IT. (Again, we omit,' thespecification of the rule for determining y's description,although'one should be.clear from the example.)
For (i), there is one such type restriction - [(Ey:Donkey). . Own.
u,y1-. The description of the existentially quantified 7 is.7.
(ii) X(u) [rtY: fcliT Own u,y]
a
That is, it is a function which, given a u, returns a donkey that
u Owns if.0 owns a donkey. For a given x then, ly: Donkey y*
Own x,y is a donkey that x owns. - Substituting into (i) yields
(iii) (qx:X(u:Man)[(Ey7.ponkey) Own u,y]). Beat x, tly:, Donkey y & Own x,y
Notice. that- this rule' is independent of how the type
containing the existential has been quantified. Thus, in
12No. Some man who owns a donkey beats it.'13. Which man who owns adonkey beats-it?
soarAv .444V 28 -
j
/
Anaphora and Logical Form
the antecedent' of "it" is the :donkey 1-obtained by applying
function (i14) to the quantified-variable associated with "man".sz !
V. Discussibn
The examples of Section IV were designed to illustrate the
feasibility of derving possible antecedents fo.r anaphoric.
expressions directly frog an appropriately structured logical..-
representation. oti.ce: that basic-to this representation is an''
adeciaati indication' of the scope of . logical operators
A 4.
o 1_,/-
quanti iers, conjunction and negation - for otherwise, we could.._) .
/pot',deal correctly with antecedents ar.ding from exiieentials,...,...
. . .. . .
"(e.g., the examples in Sections IV.A and IV.D). Also basic isI 11
tte recognition and correct attachment of aodifiets. - relative
clauses, prepositibnal, phrases,. prenomina,l alodifiert, etc: -1
. ._
. .
necessary for correctlyhandling."one" anaphora (e.g., Section
,I .B). Taken together, these impose the. requirp5nt of
pre-processor for mapping surface strings onto logital forms at
least as powerful as tht of the LUNAR system "[Woods et,
1972]; It' follow'S that much of the burden of antecedent
identification .is actually b4ing upon this pre-processar, (:
given the need for an appropriate logical form before our
approach can be applied.
It should also be clear that catat we are describing in this
paper iS essentially a competence model tor,..'anaphdr resolution:
In its crudest implementation, one would first generate a set of
- 29
Anaphora and Logical Form
possible antecedents, and then _test each of these' by plausible
.reasoning.I
using general world knowledge. Of course, we are not
seriously proposing such a generate and test implehentation.
There are a variety of heuristics that can be invoked to aid the
choice of a most probable antecedent, and any performance model
must make use of such knowledge. (Heuristics fot assigning
antecedents have. -tiden proposed throughout the linguistics,
psychology and litetatures. See [Nash-Webber, 1977] for a'
discussion of many of them.) ,Nevertheless, even a performance
model- must have the ability to determine the space of possible
'alternatives from which such heuristics are to make their choice.
Since some of these alternatives may not be present explicitly,
it is here that The approach of this paper becomes relevant.-
VI. Fur th.er Pro, lems
Ass.thisspaper.is necessarily brief, we do not have the space
to discuss at 'length such interesting issues as the effect's of
negation or various opaquecontextson the kirids of antecedents
evoked. These are discussed in [Nash-Webber, forthcoming]. We
will, however, mention one such issue -- the problem of
-existential noun phrases in negative contexts.
The scope of negation is inherently ambiguous, and ag with°
quantifiers, different scope interpretAtions yield different
antecedents. 'Moisover, some inter4dtations may 'ydeld no
-
antecedent at all. For example, .-we know 'that in positive
- 30
.33
ti
Aniaphora and Logical Form
context, an existential noun phrase will always result in an110
entity which can be described in terms of that context.' So; if
John married a Swedish girl, "she" can defer to the Swedish girl
that John married. However, in a negative context, "she" may
refer.to other entities,, depending on how thp scope of negation is
interpreted.
Negated Verb......... ....i . '
14a. John didn`t-marry a Swedish girl.b.. He lived with her for three years.
...>
("she" = the Swedish girl John was involved with)
Negated Modifier
15a. John didn't marry a Swedish girl.b. She was from Denmark.
("she" = the girl John married)
Negated Maih Descriptor
16a. Jahn didn't marry a Swedish girl.b.'She was at least 15 years his senior.
("she" = the - Swedish female that John married)
Negatd Proposition
17a. John didn't catch a trout..b. *He ate it for dinner.
Again, we would 'want to postulate a neutral. initial.--,.
< .1
representation for negation, one that Might be vague, but would
nevertheleSS be true. Only when we were required to - e:g., in.\
Order to ceaolve an anaphoric expression - would -we then attempt,..._---------
. to make a commitment to the scope of negation. (Note that 'a
belief context poses much the same- problem as negation, i.e. that
of determining the scope of belief. For example, in
- 31 -
Anaphora and 'Logical Form
,.
1..a. JOhn thought he married wSwJdish girl,- - b. but she wes really frOm DenmaL4.
,t. .as in E
iample 15 above, "she" s the girl that John rlrried.,
' -Here again, it 1,5 oply the modifier "Swedish', that should be.',...). .
-taken as- falling within. the scope of belief.).
As we me,ltioned e-arlier/ before_ one can fix on a parijicular
Gepcescntation 'language, one t provide for mass c ncepts; for
tense; for guantifier5 other than -universals and e i tentials;
for facts, ,events, states or acts; and for-generics, am ng other
thinn:3, as the(following examples illustrate.
19a. when Johnspilledbeer-on'the sofa;his dog -licked up. c \
-( "it" the - specific quantify of.beer John spilled onthe sofa), 4
20. John drinks beer because it tastes_good;-("it" = beer) __
. 21a. Many linguist's smoke,. b, although they know it causes cancer.-( "they" = the linguists cT/Eo smoke;
"it" = smoking')
. 22a. .Fe.w linguists smok,e,b. since .they know it causes cancer.
-'("thec7-1=-Tinguists,)
23a.. A beagle smiled at me yesterday.b. They are very friendly dogs.
=("theTT-= the generic clas5 of_beagles)4
,24a. John dUnked Mary's braids in the inkwell..
B- pause
1.
it, made, her -cry, he apologized for doing it2" a-.
-the event of John'sdunking Mary's braidsin the inkwell;
"it2
" = the'act of. dunking Marys braids in the inkwell)
4d
Finally, although:we have'indicated the need,for2plausible
inference fWchoosing,an appropriate candidate from a set ofP
1
- 32
.
).
, Anaphora afi,4 Logical Form. eot
,
pciss,ile antecedents (e.g, . Section IV.B.), it is also the case
that such it-erenciri\may be neededlow,/ e to derive rpossble
antecedents. Tnat is, not possible antecedents are
-structurally derivable.
025a. Yesterday I saw a'couple in the p k.
-e --b. He was wearingTorts and she h on a dashiki.
Clearly, what is required is somegort of general knowledge -of/:
ttl'e form: "A couple' usually consists of two individuals; one
'male and one female."
Although we can see no a' nriori reasons why a formal
approach could not accommodate the use o.f Plausible reasoning in
the derivation of possible, antecedents, we have chosen not
explore these isspes- in this paper. Rather, our intention in
this work is to first- determine just. how far an essentialry
syntactic approach can be pushed.
In thi s connection, no4ice that our-treatment of all of the
examples of Section IV has a decidedly syntactic character:
descriptions of those entities ,,roposed as possible, antecedents
ar-e- either explicitly present in some formula of, the MRL, or can
be derived from,such a formula by appropriate local operations on-
its structure",' independent of the availability of general world
knowledge. The determination of poSsible antecede1 ts. based on
such purely syntactic considerations and title formulation of/
4
, 4-
designfonstraints on 'MRLs. to facilitate this process best
--4,,,
describe the objectives of this paper./ '.4,
33 -
.
) . - a
7Anaphara and Logical Form
#
ReferencesA ). ,
s ,)7 [1] bresnan, J. (1971)
,.
,"A -Note 911the Notion 'Identity of .Sense Anaphou'd,'Linguistic :
Inquiry, Vol. II(-4), pp. 589-597. .L
[2.1 Charniak,:E. and.Y. Wilks leds.j (1976fComputational"Semantics. Nosth-Holland Pubrishing.Co.,,Amsterdam.
[3] Chomsky, N. (1975)-
Reflections on Language. Pantheon Books, New,York.f
. [4] CODASYL (1971)- ~
Data Bas,e,Task Group of CODASYL Programming', Language CommatteeReport,.ACM.Publicafions, New York.
[5] Codd,'T. (19(0)!` "A Relational Model of Data for Large Shared Data Banks," CACM
13, No; 6.
[6] Deutsch; fr.-%(1975)"EstablishingContext in TZSk Oriented Dialogs," American Journalof Computational Linguistics, No. 4.
[7] Hendrix, G. -(1975)"Expanding the Utility .;of Semantic Networks through'Partitioning," Advanced Papexs of 4IJCAI, Tbilisi, USSR, 3-B,
-September, 1975.
[8] Langacker, R.' (1966) A",Pronominalpiiation and the Chain of Command," in Modern Studiesin English, Reibel and SchaneledS-.) r Prentice-Tiff-5ubri-7-14:eWJersey;
7-
'[9] "Nash-Webber, B.L. 41977)"Anaphora:: A Cross-discidlinary Survey" , Technical Rebort CSR-31.Center for the Studx of Reading, University of Illinois and Bolt'Beranek and Newman,-Inc.
[10] Nash-Webber, BL.(forthcoming)Forthcoming doctoral dissertation, Department of AppliedMathematids, Harvard University.
- [11] Norman, D. & D. Rumelhart (1975)Explorations in Cogniiion,M.H. Freeman & Company, San Francisco.
inhart,,, T. (1976)° lhe -tactic Domainawfof Anaphora," unpublished doctoral
disser tion, .MIT 'Department of Foreign 4iteratures andLingui tics,
4
0
- 34 J. 0
7
16
e
[13] Ross, J.R. (1967)
Anaphoraand Logical Form
"C nstraints on Variables in. ,Syntax," unpublished doctoraldi_.serEation, MIT Department, of Foreign -LiteKatures andLinguistics. (Available from the Indiana University LinguistidsClub, Brominlgton, Indiana.)
[14] Sag., I. (1976)A .
"Deletion and Logical Form," unpublished doctoral dissertation,MIT Department Of Foreign Literatures and Linguistics.
115] SandeWall, E.J. (1971),
i"Representing Natural Language In formation in 11
.10t
,PeedicateCalculus," in Machine Intelligence 6, Edinburgh University Press.
[16] SchanW,,R. (ed.) (1975)Conceptual Information Processing, North-Holland Publishing Co.,Amsteraam.
[17] Schubert, L. (1976)"Extending the Expressive Power of Semantic Networks," ArtificialIntelligence 7(2)., pp. 163-198.
[18],Simmons, R.F. (19733,"Semantic Networks: Computation and Use for Understanding EnglishSentendes," in Computer Models of Thought and' Language, Schankand Colby (eds.), W.H. Freeman & Co., San Francisco.
[19] Simmons, R!F. and,B.C. Bruce (197.1)"Some Relations between Predicate Calculus and Semantic NetRepresentation of.Discourse," in Proc. 2nd IJCAI, London.
[20] wilks, Y. (1975)"A. Preferential, PaXsteZn-sieeking Semantics forNatural'Language,"Artificial Intelligence 6,. pp. 53-74.
[21] Williams, . (1977)"Discourse and Logical Form", Linguistic Inquiry 8(l), pp.
'101-140. . A0
[221 Woods,' W.&. (1975).1 "What's in a Link: Foundations for Semantic Networks," in Bobrowf-and Collins (eds.), Representation and Understanding: Studies inCognitive Science. New. -York: Asa-der-6TE Press.
'[23hrWbods, W.A., R.M. Kaplan and B.L. Nash-Webber (1972)"The Lunar Sciences Natural Language Information System: FinalRepor BBN eport, 2378, Bolt Beranek and Newman Inc.,Cambr ge M1? (Av ilable\from NTIS as N72-23155).
p
r
- 35
3 8 .
..
ti
CENTER FOR THE STUAY OF,43.EApING
. No. 1:
TECHNICAL REPORTS
Halff, H. M. Graphical Evaluation. of Hierarchical ClusteringSchemes, October 1975.
,
No. 2: Spiro., R.-J. Inferential Reconstruction inliemory for ConnectedDiscourse, October-1975.
No. 3: Goetz, E. T. Sentences in Lists and in Connected Discourse,November 1975.
No 4: Alessi, S. M., Anderson, T. H., & iddle, W. B. Hafidware,and ,4
'Software Considerations in Computer Based Course Management,November 1975; /
No. 5: Schallert, D. L. ,IMproving Mem ry for Prose: The RelationshipBetween Depth; of Processi g_agi Context, November 1975, .
/ .
No. 6: Anderson, R. C., Goetz, E. T. Pichert, J. W., & Halff, H. M.) Two Faces' of the Conceotual Peg Hypothesis, January 1976.
,-r,1No . Ortony, A. Names, Descriptions, and Pragmatic;, FebrUary 1976.
r
No. 8: Mason, J, M. Questioning the'Notign of Independent ProcessingStages in Reading, February 106.
orNo. 9: Siegel, M. A. %eacher Behaviors and Curriculum Packages:
-Implications for Research and Teacher Education, April 1976.
4 a
No. 10 Anderson, R. C., Pichert, Jc W., Goetz, E".. T., Schallert, D. L.,Stevens, K. V., & TroIlip, S. R.' Instantiation of General
je
No. 11:
No. 3:2.:v.
Terms, March 1976. - ,
Armbruster, B. B. Learning Principle's from Prose: A CognitiveApproach Based on Schema Theory, July 1976.
AnAerson, R. C., Reynolds, R. E., Schallert, D. L., & Goetz,E. T.
Frameworks for Comprehending Discourse, July 1976.
No. 11: Rubin, A.D,,Bruce, B. C., & Brown, J. S. A Process-OrientedLanguage for Describing Aspects prReading Comprehension,
.November 1976
a,
No. 14: Pichert, J. W., & Anderson, R. C. Taking Different.Perspectives'.'on a ttory, November 1976. ,
14. 15:_.Schwartz, R. M. ,Strategic.Procepses in Beginning Reading.--November 1976.
A 39.4*
t
?:,z)
. Ilk
No. les:. .Jenkinq, J. R., & Fany, B. Curriculum Biaes in Reading,,KbhievemeDt Tests, November 1976.
.. .
No. 17: Asher, S. R., Hymel, S., & Wigfield, A. Children's Co5thehensionof High- and Low-Interest Material and a'CoMMason of TwoCloze Scoring Methods, November l'876` . ,-
45 ,.. ,,-', No. 18 :' Brown, A. L., Smiley, S. S., Dv011. D., Townsend, M. A. R.,
& Lawton, S. C. Ingilaitn 644,Thematic.Idea in Children'sComprehension and 11-tention of'Stories, December 1976.
No. 19: Kleiman, G. M. ThePrelinguistic Cognitive Basis*of Children'sCommunicative Intentions, February 1977.
. -No.20: Kleiman, G. M. The Effect-of Prel>iousiC ontex t on Reading
Individual Worlds, rebruary 1977.,.,
No. 21: Kane, J. H., & Anderson, R. C. Detof Processing andInterference Effects in the Learning and Rememberinfr of
No., 22:
Sentencef, February 1977..-
Brown, A. L., &,Campione, J. C. Memory Strategies' in LearningTraining Children to Study SirategicallyOlarch;1977.
No. 23: Smiley, S: S., Oakley, D.Brown, A. L. Recall
D., Worthen, D., Campione, J. C., 6;of.Thematically Relevant Material,by
Adolescent Good and Poor Readers as a Function of Writ enVersus Oral Presentation, March 1977.
No. 24: Anderson, R. C.i Spiro, R. J.011 Anderson, M. Schematd asScaffolding for the .Representation of Information inConnected Discourse, Marti 1977.
Pany, D., & Jenkins, J. R. Learning Word Meanings: A Comparison 'of Instructional Procedtxses and Effects on Measures of ReadingComprehension with Learning Disabled Students$ March 1977.
Armbruster, B.3., Stevensor.R. J., / Rosenshine, B. AnalyzingContent Coverage and Emphasis:'t.9A Study of Three Curriculaand Two-Tests, March 1977.
No. 25:
No.z26:
No. 27: Ortony, A.,-Reyndlds,- R. E., & Arter, J. A. Metaphor: Theoretical_and Emplrlcal Research, March 1977%
No. 28:- Ortony, A. 'Remembering and Understanding Jaberwocky an mall-Talk,Mardh
4o, 29: Scbailert, D. L., Kleiman, G. M., & Rubin, A. D. Analysis of 'Differences Between Oral and ritten Language, Aprik1977.
Goetz, E. T., & Osborn, J. Procedures/tor Sampling TextSand Tasksin Kindergarten-through Eighih:Grade, April 1977.
er-40
A A
No. 3,1:. Nash-Ciebber, B. Anaphora: A Cross - Disciplinary Survey, April 1977.
No. 32: Adams, M. J.; & Collins, A. A Sc hema-Theoretic View of Reading
Comprehension, April 1977. t
No. 33: Huggins, A. W. F. Syntactic Aspects of Reading Comprehension, April
° 1977.
No. 34: Bruce, B. C. Plans and Social Action, April 1977.
Nor.'55°: Rubin, A. D. Comprehension Processes in Oral and Written Language,
April 1977.4)
NO. 36: Nash-Webbbr h B. B & Reiter h. Anaphora-Meaning Representations for Natural
No. 37: Adams,4. J. ,Failures to Comprehend: A
and. Logical pnFormalLanguage; April 1977.
QuestiOn of Balance, April 1977.
,
No: 38: Woods, W. A. Multiple Theory Formation in High-Level Perception,April 1977. °
No. 39:'"k6
Nicke,rson, R. S.; & Adams, M. J. Usea:of ontekt in Speech
Understandingand Reading, April 1977.
No.; 40.: Brown; U. S.N Collins, A.' Model-Based Versus Text-Based Reasoning,
April 1977.
-No..41:
°
No. 42:
No. 43:
47 411.4t-
Anderson, R.'C., & Pichert, J. W. Recallof PreviousZy,UnrecillableInformation Following a.Shift in Perspe#ive,,Aptil 1977.
Mason, J., Osborn, J., &.Rosenshine, B. ,A Consideration of SkillHierarchy Approaches -to the TeaChing of Reading, April 1977.
,
4.14-
Collins, A., Brown, A. L., Morgan,',..1 L., & Brewer, W. F. The
Analysis of Reading-Tasks and Texts, April'1977. iqk
' NO. 44: McClure, E. Aspect's of Code-Switching Oiscouese of Bilingual'.0,psMexican-American ChildAp., Apri11977.
No: 45: Schwartz, R. M. Relation of Context Utilization and Orthograribic opAutomaticity in Word Identification, May 1g77.
Anderson, R.. C., Stevens, K. C., Shifrin, Z., & Osborn, J.'Instantiation ofy Word Meaning in Children, May 1977.
.
No. 47: Brown, A. L. Knowirig When, Where, and How to RemembOr- A Problem
j.
v of Metacognition, June 1977..
o
No. 46:
,,.. .-No. 48: Brown, A. .L., & DeLoache, J.S. Skills,_ Plans, and Self-Regulation,'
I 4.
July 1977.. *,..-
No. 4g: Goetz, E. T. Inferences in theComprehension of and Memory or
Text,-July 1977.
No. Anderson, R. C. Schema-Directed Processes in Language Comprehension,
°.July 1977.
No. 51: Brown, A. h. Theories of Memory and the,Problems of Development:
/ Activity, Growth, and Knowledge, July 1977.
O
a
.7
r
If