Judging the Coherency of Discourse

American Journal

of

Computational

Linguistics

Microfiche 35 : 36

D e p a r t m e n t of I n f o x m a ti on E n g i n e e r i n g U n i v e r s i t y of I l l i n o i s at Chicago C i r c l e

B o x 4 3 4 8 , C h i c a g o 6 0 6 8 0

ABSTRACT

The component p r o p o s i t i o n s of a c o h e r e n t d i s c o u r s e exhibit a n a p h o r i c , s p a t i o - t e m p o r a l , c a u s a l and t h e m a t i c s t r u c t u r e s . Not a l l o f this s t r u c -

t u r e i s e x p l i c i t , b u t must be i n f e r r e d u s i n g a model of c o g n i t i v e know-

ledge. The o r g a n i z a t i o n of knowledge i n the model a l l o w s a bottom-up

a n a l y s i s of d i s c o u r s e . F u r m e r , knowledge i s formed i n t o s m a l l complexes

rather than i n t o t h e l a r g e m o n o l i t h i c s t r u c t u r e s found i n Scripts/Frames.

1. The S t r u c t u r e of Coherent Discourse. - -

A d i s c o u r s e i s judged c o h e r e n t i f its c o n s t i t u e n t p r o p o ' s i t i o n s a r e connected. Various types of c o h e s i v e l i n k s a r e observed i n d i s c o u r s e :

a n a p h o r i c , w a t i a l , t e m p o r a l , c a u s a l and t h e m a t i c . W e w i l l f o r m a l l y

describe t h e s t r u c t u r e of a well-formed d i s c o u r s e in terns of t h e s e

c o n n e c t i v e s .

1.1 Anaphora.

Two k i n d s of anaphora can be d i s t i n g u i s h e d . The f i r s t i s marked

by the presence of a p r o f o m ( o r by the r e p e t i t i o n of a form) :

(1) Henry travels t o o much. He i s g e t t i n g a f o r e i g n a c c e n t . Antecedents may be nominal, v e r b a l o r c l a u s a l .

term f o r the a n t e c e d e n t . For example,

(2) John p u t the car i n t o 'reverse' instead of ' d r i v e '

and h i t a w a l l . The m i s t a k e c o s t

him

$200 i n r e p a i r s .

'Mistake1 i n ( 2 ) i s an abstract c h a r a c t e r i z a t i o n of t h e g e a r s e l e c t i o n

e x p r e s s e d i n the f i r s t s e n t e n c e .

A c o n v e n t i o n a l way t o label the r e c u r r i n g a c t o r s i n d i s c o u r s e i s

a s I d r a m a t i s p e r s o n a e ' . However cohesion can r e s u l t n o t o n l y f r o m

m u l t i p l e appearances o f p e o p l e , but of any concept, as i n ( 2 ) .

1 . 2 Spatio-temporal and Causal Cohnectives.

Space, time and cause give coherency to a s e t of p r o p o s i t i o n s .

( 3 ) The King was

in

the c o u n t i n g house, c o u n t i n g o u t h i s money. The Queen was i n the p a r l o u r , eating bread and honey.

The actions i n ( 3 ) a r e set i n d i f f e r e n t rooms, but of the same ' p a l a c e 1 .

( 4 ) A f t e r Richard t a l k e d t o the r e p o r t e r , he went t o lunch.

The temporal s e m e n c e of events i n ( 4 ) i s expressed by ' a f t e r 1 .

(5) John e a t s g a r l i c . Martha a v o i d s him.

To non-aficionados garlic i s known o n l y

for

i t s aroma, detection of ~ h i c h causes e v a s i v e a c t i o n .

Cause, i l l u s t r a t e d i n (5) i s a n important d i s c o u r s e connective.

N o t e however, t h a t this i s an e t h n o c e n t r i c view; i n akher cultures a

different position may have t o be t a k e n , f o r example, a t e l e o l o g i c a l world view (White : 1975)

.

!Ibis dimension of discourse s k r u c t u r e is termed i t s ' p l o t ' s t r u c t u r e . 3.. 3 Thematicity.

6 Dino Frances drowned today i n Middle Branch Resevoir a f t e r r e s c u i n g h i s son Dino Jr. who had fallen i n t o t h e water while on a f i s h i n g t r i p .

i s a new s t o r y from the New York T i m e s , w i t h a theme o f , say, ' t r a g e d y ' . Discourse may have more t h a n one theme, b u t t h e s e should n o t c o n f l i c t .

( 7 ) E a t i n g t h e fish made Gerry s i c k . He had measles i n May. I n ( 7 ) we have an i n c o h e r e n t s t r u c t u r e . The p r o p o s i t i o n 'Gerry s i c k '

belongs b o t h to a t o p i c 'food-poisoning' and t o a biography of i l l n e s s e s .

The a n a l y s i s of f a i r y - t a l e s by Lakoff (1972) s u g g e s t s that d i s c o u r s e has

a s t r i c t l y t r e e - l i k e thematic o r g a n i z a t i o n .

I t i s concluded t h a t the p r o p o s i t i o n s of a c o h e r e n t d i s c o u r s e a r e

connected e i t h e r by c o r e f e r e n c e O r ( p r e f e r a b l y ) c a u s a l l y , and that it

has a single theme (which m a y be t h e r o o t of a t r e e of themes).

2 , The Role o f I n f e r e n c e .

Not a l l of d i s c o u r s e s t r u c t u r e i s o v e r t l y s t a t e d ; d i s c o u r s e i s h i g h l y

e l l i p t i c . I n ( 4 ) the d i s c o u r s e connective ' a f t e r ' i s p r e s e n t t o mark a

temporal sequence, b u t i n ( 5 ) t h e r e i s no r e a l i z a t i o n of the c a u s a l r e l a t i o n

between t h e two p r o p o s i t i o n s . Normally one assumes t h a t a d i s c o u r s e i s

coherent; hence ( 3 ) i s most a c c e p t a b l e i f t h e rooms a r e taken a s being with-

i n the same h a b i t a t i o n . E v i d e n t l y a r e a d e r must i n f e r o m i t t e d s t r u c t u r e . The i n f e r e n c e s axe made f r o m h i s c o g n i t i v e s t o r e of world knowledge.

There i s much d i s c u s s i o n a t p r e s e n t about i n f e r e n c e a s p a r t of under-

s t a n d i n g . To make i n f e r e n c e s i s easy; t h e problem i s t o make t h e r i g h t

ones. I t h e l p s t o have a goal. I t i s suggested that d i s c o u r s e can be

3. Mechanisms of I n f e r e n c e .

A model of cognitive knowledge

--

a n encyclopedia

--

s h o u l d be capable of making the i n f e r e n c e s n e c e s s a r y t o form an o p i n i o n about

the coherency of a d i s c o u r s e . The p r e s e n t e n c y c l o p e d i a o r i g i n a t e d w i t h Hays ( 1 9 7 3 ) ; a fuller d e s c r i p t i o n can be found i n Phillips ( 1 9 7 5 ) . It

is implemented

as

a d i r e c t e d graph. Labeled nodes c h a r a c t e r i z e concepts

and labele'd a r c s r e l a t i o n s between c o n c e p t s .

P r o p o s i t i o n s have a s t r u c t u r e of c a s e - r e l a t e d c o n c e p t s , based on

F i l l m o r e (1989). This i s our ' s y n t a g m a t i c ' o r g a n i z a t i o n of knowledge.

AS p r o p o s i t i o n s a r e e s s e n t i a l l y the b u i l d i n g b l o c k s of d i s c o u r s e , w e will n o t d w e l l on t h e i r s t r u c t u r e h e r e .

3 . 1 Anaphora.

I f the dependent i s a profom then part of u n d e r s t a n d i n g i s t o determine the c o r r e c t a n t e c e d e n t . There are s y n t a c t i c constraints

(Langacker: 1969) which serve t o narrow down c h o i c e s f o r anteaedents and

t o g i v e an order of preference. The chosen a n t e c e d e n t w i l l be the f i r s t

t h a t , w h e n s u b s t i t u t e d f o r t h e proform, produces a m e a n i n g f u l p r o p o s i t i o n

that is c o h e r e n t i n c o n t e x t .

A meaningful p r o p o s i t i o n is one t h a t h a s a c o u n t e r p a r t i n the ency-

clopedia. The counterpart m a y be the self-same p r o p o s i t i o n , o r m o r e

l i k e l y . a g e n e r a l i z e d proposition (hereafter a GP]. For example,

rather than 'Joan drink m i l k ' , w e would e x p e c t t o find 'animal imbibe l i q u i d ' .

How are GPs found? All concepts belong t o p a r t i a l l y o r d e r e d

taxonomic structures i n t h e encyg-edia (our ' p a r a d i g m a t i c ' organiz-

a t i o n of concepts). _From any c o n c e p t i t i s possible t o f o l l o w para-

uent

of

a proposition,

An

i n t e r s e c t i o n

of

paradigmatic

paws o r i g i n -

a t i n g from each c o n c e p t

i n

a d i s c o u r s e p r o p o s i t i o n ( h e r e a f t e r a D P ) ,

t a k i n g a c c o u n t ~ f s y n t a g m a t i c s t r u c t u r e , gives a GP. If there i s no

such

i n t e r s e c t i o n , then the Dl? is not c o n s i s t e n t with e n c y c l o p e d i c

knowledge.

Abstract

terms

can be defined by complexes o f GPs, each

having

sufficient c o n c e p t u a l content t o d e f i n e s i t u a t i o n s i n which they apply.

For example, a d e f i n i t i o n o f ' m i s t a k e ' must be such t h a t it applies to

part of

the first

s e n t e n c e i n ( 2 )

.

3.2 Space, T i m e

and

Cause.

To

i n f e r

oxni-tted spatio-temporal and causal r e l a t i o n s ( t e m e d

' d i s c u r s i v e ' r e l a t i o n s i n t h e e n c y c l o p e d i a ) , i t is also necessary to

l o c a t e GPs. T h e e n c y c l o p e d i a , o f course, i n c l u d e s t h e s e r e l a t i o n s , but

between GPs. ~ c h e m a t i c a l l y , from a discourse p r o p o s i t i o n P

we

can 1

l o c a t e P a GP, i n t h e manner o u t l i n e d above. P may have a discursive

2 ' 2

r e l a t i o n R t o a n o t h e r G P ,

P 3

.

A p r o p o s i t i o n P 4 ' a p a r t i c u l a r i z e d version of Pj, and t h e r e l a t i o n R , between P and P can be added t o t h e

1 4 '

discourse, f i g u r e 1.

Often P4 will be a p r o p a s i t i o n a l r e a d y s t a t e d i n t h e d i s c c ~ n z o e ; merely t h e r e l a t i o n need be i n f e r r e d t o augment t h e p l o t structure, It may,

however, be necessary to, i n f e r a chain of p r o p o s i t i o n s t o l i n k the

o r i g i n a l DPs. The q u e s t i o n a r i s e s whether there i s a l i m i t on t h e number of p r o p o s i t i d n s i n a ' s e n s i b l e ' i n f e r r e d path, I n t u i t i v e l y

there i s , but at present we have no f o r m a l i n s i g h t . 3 . 3 Thematicity.

A theme i s a complex of G P s , s t r u c t u r a l l y i n d i s t i n g u i s h a b l e f r o m

that used i n c h a r a c t e r i z i n g a b s t r a c t terms l i k e ' m i s t a k e 1 , The potential presence of a theme i s d e t e c t e d i n the p r o c e s s of s e e k i n g G P s f o r D P s .

All G P s , whether o r n o t they are part o f a t h e m a t i c d e f i n i t i o n , can be l o c a t e d by p a r a d i g m a t i c s e a r c h e s ; some GPs have additional structure

i n d i c a t i n g t h a t they are components of themes. Tt i s n o t s u f f i c i e n t t o

e s t a b l i ~ h a theme f o r d i s c o u r s e by s e p a r a t e l y f i n d i n g D P s that correspond

t o all the GPs of a theme. The thematic definition and t h e r e l e v a n t

?art of the discourse m u s t be t e s t e d h o l i s t i c a l l y t o e n s u r e that t h e

c o r r e c t c o r e f e r e n t i a l i t i e s exist among the p r o p o s i t i o n s ,

3l4 Overview of Inference, _{- -}

There are two b a s i c p r o c e s s e s u n d e r l y i n g i n f e r e n c e . F i r s t there is the process of l o c a t i n g a G P given a DP. This i s implemented essen-

t i a l l y by a b r e a d t h - f i r s t search t h r o u g h the paradigmat~c structure o f

the encyclopedia. Secondly there i s the p r o c e s s of matching a complex

of p r o p o s i l 5 o n s i n discourse a g a i n s t an e n c y c l o p e d i c complex. The

latter p r o c e s s is q u a l i t a t i v e l y different a s it involves t e s t s f o r co-

reference that the former does not.

'Frames1 (Minsky: 1975). Adding t o the expanding terminology, o u r v e r s i o n known

'

m e t a l i n g u a l d e f i n i t i o n s ' .

Metalingual d e f i n i t i o n s s e r v e t o d e f i n e a b s t r a c t terms ( I m i s t a k e 1 ) ,

themes ( ' t r a g e d y 1 ) and p l a n s (used by Furugori ( 1 9 7 4 ) i n his r o b o t p l a n n e r ) . The d i s t i n c t i o n s a r e more t e r m i n o l o g i c a l t h a n s u b s t a n t i v e ,

t h e i r f u n c t i o n s are i n t e r c h a n g a b l e ; i r l o t h e r c a n t e x t s a p l a n could be a

theme, a theme an a b s t r a c t term, e t c .

When an a b s t r a c t concept has a m e t a l i n g u a l d e f i n i t i o n , a matching d i s c o u r s e may be r e w r i t t e n i n terms of that concept. For example, 'buy'

h a s such a d e f i n i t i o n , s a y ' p e r s o n g i v e s o b j e c t t o pexson

,

person

1 2 2

g i v e s money t o person To p r o p e r l y make t h e t r a n s d u c t i o n t o ' p e r s a n

1 - 2

buys o b j e c t from p e r s o n

,

t h e r e must be a c a s e frame f o r 'buy' l i n k e d

1

t o concepts i n i t s d e f i n i t i o n . A p r o p o s i t i o n produced by a b s t r a c t i o n

i s s t r u c t u r a l l y i n d i s t i n g u i s h a b l e from a p r o p o s i t i o n t h a t was i n the

i g i n a l d i s c o u r s e , and can be s u b j e c t encyclopedic p r o c e s s ,

i n c l u d i n g f u r t h e r a b s t r a c t i o n . Conversely, i f a p r o p o s i t i o n c o n t a i n s

concept having a m e t a l i n g u a l d e f i n i t i o n , then t h e p r o p o s i t i o n can

be decomposed i n t o a complex of p r o p o s i t i o n s p a t t e r n e d on t h e d e f i n i t i o n ,

4. An Example.

A schematic a n a l y s i s of ( 6 ) shows t h e i n f e r e n c e system i n o p e x a t i o n ,

r e s u l t i n s i n a structure t h a t s a t i s f i e s t h e c r i t e r i a of coherence,

A t each s t e p w e w i l l i n d i c a t e t h e encyclopedic knowledge used i n

the i n f e r e n c e , and t h e c u r r e n t s t a t e of t h e d i s c o u r s e . The o r i g i n a l

Step 0 , Initial S t a t e .

F a t h e r

0

drowns

a

Father

rescue son

Son in

w a t e r

Son falls

- . - - - I - - - - . . C I C C - - C I I I - - - . L I I I I I I L C C C C - C C - - -

Step 1, Fall causes injury.

Father d r o w n s

e

Father

rescue son

Y

watex

CAUSE

Son falls

Son

Step 2 . Injury causes inability to a c t ,

Father

drowns

0

Father rescue

s o n

h

Son in

water

f m U s E

-+

CAUSE

Son Son Son not able

falls injured to act

- - - --

Step 3 , In water and not able to act causes rescue.

e

_drowns

Father

res cue

Son Son Son not able

f a l l s in j w e d t o a c t

Conjunction is i n d i c a t e d by P a r t - w h o l e r e l a t i o n s . N o t e t h a t a link to

one of the o r i g i n a l propositions has been established.

Step 4 . To rescue someone w h o is in water it may be necessary to be in w a t e r .

Father

*owns

CAUSE

-.,

_Father i n

h= + . v - t a -

< .

CAUSE T-w

'II

'.

Father

\, res cue

&,

son

%A\

w a t e r

s

CAUSE CAUSE

----,(--J - - ---- -=- --- --- - -

Son Son Son n o t able

falls injured ta act

Step 5. A c t i n g can make you we-.

Father

e

d r o w n s

w a t e r

'.

CAUSE CAUSE

Son Son _{Son not able}

Step 6. If weary then unable to act.

Father

a

drowns

CAUSE Father in

CAUSE _{- -}

_--a

CAUSE

4

3

,

Father rescue

F a t h e r Father lot

weary able t o a c t

water

f

, CAUSE _{-- -}

-

-> A& .

Son Son Son not able

falls i n j u r e d to act

Step 7. I f i n water and not able to act t h e n drown.

CAUSE

Father Father n o t rescue w e a r y able t o act

Son f a l L s

Son

injured

Son not able t o act

e n t i a l i t y c o n d i t i o n s p r e v e n t 'son i n water1 and ' F a t h e r not a b l e t o act1 c o n j o i n i n g t o s a t i s f y the c o n d i t i o n s on this i - n f e r e n c e .

Note that the a n t e c e d e n t c o n d i t i o n on t h i s inference i s the same as at step 3 . Both resultant situations are p o s s i b l e , and axe noted. The system can select either. However, the wrong c h o i c e does n o t l e a d

t o a connected s t r u c t u r e , and a back up t o the a l t e r n a t i v e has t o be

m a d e .

The d i s c o u r s e now h a s an i n f e r r e d causal s t r u c t u r e c o n n e c t i n g a l l the original p r o p o s i t i o n s .

From a t h e m a t i c a n a l y s i s of drowning s t o r i e s i n g e n e r a l ( P h i l l i p s : 1975), t h e common theme can be described as ' g i v i n g a cause for t h e

person being in a e w a t e r , and g i v i n g a cause for t h e v i c t i m n o t b e i n g

able t o a c t (thereby n o t b e i n g able t o s a v e himself)'. T h i s theme fits

t h e d i s c o u r s e by virtue of p r o p o s i t i o n s

a

and

0 ,

which stand i n causal r e l a t i o n s t o 'being i n t h e water' and ' n o t a b l e t o a c t ' f o r the v i c t i m , The theme ' t r a g e d y ' is defined as 'someone does something good

and d i e s as a result of this actiont. The father's rescue of h i s son and subsequent demise satisfy this theme ( @ and ) For t h e s t o r y to

be coherent, these themes must n o t o v e r l a p ; i n fact w e s e e that the

l drowning' theme is properly contained by

'

tragedyt

.

5. Discussion.

The analysis is s o o r g a n i z e d t h a t the themes are d e t e r m i n e d i n a b o t t o m up manner, as are a l l generalized f a c t s used i n the a n a l y s i s .

Though n o t p r e s e n t l y implemented, it s h o u l d be p o s s i b l e t o use p o t e n t i a l themes, ones for which only some component p r o p o s i t i o n s have been found,

The complexes of propos'i t i o n s

,

i n rnetalingual d e f i n i t i o n s of themes

and elsewhere, are r e a l l y n o t t h a t complex. The ones i n t h e example

c o n t a i n only a few p r o p o s i t i o n s . Each h a s only the e s s e n t i a l s of the

s i t u a t i o n . The f i n a l s t r u c t u r e a r i s e s from many s m a l l p i e c e s of

knowledge r a t h e r t h a n from one m o n o l i t h i c aggregate. This s e e m s t o b e

a more n a t u r a l o r g a n i z a t i o n , a s each ~f t h e s i m p l e r s t r u c t u r e s can be

f r e e l y a p p l i e d i n many c o n t e x t s , r a t h e r t h a n being bound t o one s i t u a t i o n .

The d i s c o u r s e judgement i s r e l a t i v e t o t h e knowledge of t h e h e a r e r .

Whether t h e i n f e r e n c e s a r e t h o s e i n t e n d e d by the a u t h o r i s a n o t h e r

q u e s t i o n . I d e a l l y t h e y should b e r o r d i f f e r e n c e s s h o u l d b e unimportant.

A misleading i n f e r e n c e i n d i c a t e s poor w r i t i n g by t h e a u t h o r ; h e h a s

misjudged the knowledge of his audience.

D i r e c t i n g i n f e r e n c e s on a d i s c o u r s e towards the g o a l of judging i t

coherent provides a normalized v e r s i o n of the d i s c o u r s e , i f the p r o c e s s i s s u c c e s s f u l . The normalized s t r u c t u r e can form the b a s i s f o r f u r t h e r p r o c ~ s s i n g : c o n t e n t a n a l y i s , s t y l i s t i c a n a l y s i s , e t c . It may a l s o

provoke various q u e s t i o n s , for example, w e could ask i f the i n f e r e n c e s were c o r r e c t ; we have t h e ' r e s c u e ' s i t u a t i o n applying t o the father, b u t

ieferences

.

?illmore, C. J. 1969. Toward a Modern Theory of Case. In R e i b e l and Schane.

Furugori, T. 1974. "A m e m o r y

model

and

simulation of

m e m o r y processes for driving a car." T e c h n i c a l Report N o . 7 7 , Department of Computer S c i e n c e , SUNY Buffalo.

Hays, D. G. 1973, Types of Processes on Cognitive Networks.

In Proceedings of the 1973

--

International Conference

-

on Computational Linguistics. Pisa.

Lakoff, G. 1972. Structural Complexity in Fairy Tales.

me Study of Man 1, 128-150.

-

--

Langacker, Ro W. 1969. On Pronominalization and t h e Chain of Command, In Reibel and Schane.

Minsky, M. 1975

.

A Framework for Representing Knowledge.

In P

.

H. Wins t o n (ed. )

,

-

The Psychology of Computer Vision,

. )

McGraw-Hill

,

MI.

Phillips, B. 1975. Topic Analysis. Unpublished Ph.D, Thesis. SUNY Buffalo.

Reibel

,

D. A. and D. A. Schane (eds

.

]

.

1969. Modern s t u d i e s

-

in English.

Re,adings

-

in Transformational Grammar. Prentice-Hall, Englewood Cliffs,

Schank, R. C. and R. P. Abelson. 1975. Scripts, Plans, and Knowledqe. In Advance Papers

--

of the Fourth ~nternational ~ o i n t

Conference

-

on WtifLcial Intelligence. IJCAI,

White; M. 1975. Abstract Definition in the Cognitive Network:

The Metaphysical Terminology of a Contemporary Millenarian

Community. Unpublished Ph.D. Thesis. SUNY Buffalo.

Wilks, Y. 1975. A Preferential, Pattern-Seeking, Semantics for