Speech
Acts as a Basis
for U n d e r s t a n d i n g
Dialogue
Coherence
by
C. Raymond P e r r a u l t and J a m e s F. Allen
Dept. of Computer Science
U n i v e r s i t y of Toronto
T o r o n t o Canada
and
P h i l i p R. C o h e n
Bolt Beranek and N e w m a n
C a m b r i d g e Mass.
i.
Introduction
Each agent
m a i n t a i n s a model of
the
world, including a model
of the m o d e l s of
other agents.
L i n g u i s t i c utterances
are
the result
of the e x e c u t i o n
of o p e r a t o r s
whose
effects are
mainly
on the
models
that the
speaker and
hearer m a i n t a i n
of
each other.
These effects are intended by
the speaker
to be produced p a r t l y
by the
hearer's
recognition
of
the
speaker's
plan.
Webster's
dictionary
defines
"coherence"
as
"the
quality
of
being
logically
integrated,
consistent,
and
intelligible".
If one
were asked whether
a
sequence
of
physical
acts
being
performed
by
an agent
was
coherent,
a
crucial factor
in the
d e c i s i o n would
be
whether
the
acts
were
perceived
as
contributing
to
the
achievement
of
an
overall
goal.
In
that
case
they
can
frequently be described briefly, by naming
the
goal
or the
procedure
executed
to
achieve it.
Once the
intended goal
has
been
conjectured,
the
sequence
can
be
d e s c r i b e d as a more
or less correct, m o r e
or less optimal attempt at the a c h i e v e m e n t
of the goal.
This view of the c o m m u n i c a t i o n process
is very
close in
spirit
to the
AustinGrice-Strawson-Searle
approach
to
illocutionary
acts,
and
indeed
was
s t r o n g l y influenced by it.
We are working
on a
theory of speech
acts based
on the
notions
of
plans,
world
models,
plan
c o n s t r u c t i o n and plan
recognition.
It is
intended
that this
theory should
answer
q u e s t i o n s such as:
One of the m a i n s t r e a m s of AI research
has
been
the study
of
problem
solving
behaviour in humans and
its simulation by
machines.
This
can be considered
as the
task of
transforming an initial
state of
the world into a goal
state by finding an
appropriate
sequence of
applications
of
o p e r a t o r s from a given set.
Each operator
has two m o d e s of
execution:
in the first
it changes
the "real
world", and
in the
second
it changes
a model
of the
real
world.
S e q u e n c e s of
these o p e r a t o r s
we
call
plans.
They
can
be
constructed,
simulated,
executed,
optimized
and
debugged.
Operators
are usually
thought
of
as achieving
certain
effects and
of
being
applicable
only
when
certain
p r e c o n d i t i o n s hold.
(i)
Under
what
circumstances
can
an
observer
believe
that
a
speaker
has
s i n c e r e l y and n o n - d e f e c t i v e l y
performed a
particular i l l o c u t i o n a r y act
in producing
utterance
for
a hearer?
The
observer
could also be the hearer or speaker.
(2)
What
changes
does
the
successful
execution
of a
speech
act
make to
the
s p e a k e r ' s model of the
hearer, and to the
h e a t e r ' s model of the speaker?
(3) H o w
is the
meaning (sense/reference)
of an u t t e r a n c e x related to the acts that
can be p e r f o r m e d in uttering x?
The effects of one agent executing his
plans m a y
be o b s e r v a b l e by
other agents,
who,
assuming
that
these
plans
were
produced
by
the
first
agent's
plan
c o n s t r u c t i o n algorithms, may
try to infer
the plan being executed
from the observed
changes to the world.
The fact that this
inferencing m a y
be intended by
the first
agent underlies human c o m m u n i c a t i o n .
must
A theory
specify
of speech acts based on plans
at least the following:
(i)
A P l a n n i n g System:
a language
for
d e s c r i b i n g states of the world, a l a n g u a g e
for
describing operators
and
algorithms
for plan c o n s t r u c t i o n
and plan inference.
S e m a n t i c s for the l a n g u a g e s should also be
given.
(2)
Definitions
of
speech
acts
as
operators
in the
planning system.
What
are
their
effects?
When
are
they
applicable?
How
can they be
realized in
words?
* This
research was supported in part by
the National Research Council of Canada.
125
To make
possible a
first attempt
at
such
a
theory we
have
imposed
several
r e s t r i c t i o n s on the system to be modelled.
a n a l y s i s of indirect speech
acts (such as
"Can
you pass
the
salt?") utterances
which
appear
to
result
from
one
illocutionary
act
but
can
be
used
to
p e r f o r m another.
(I) Any agent Al's
model of another agent
A2 is defined in terms
of "facts" that A1
b e l i e v e s A2
believes, and
goals that
A1
b e l i e v e s A2 is attempting
to achieve.
We
are not
attempting to model obligations,
feelings ~ etc.
Section 2 of this
paper o u t l i n e s some
requirements
on
the
models
which
the
v a r i o u s agents
m u s t have
of each
other.
Section 3 d e s c r i b e s the planning o p e r a t o r s
for REQUEST
and INFORM, and how
they can
be used
to g e n e r a t e
plans which
include
assertions, imperatives, and several types
of questions.
(2) The only
speech acts we try
to model
are some
that appear
to be
d e f i n a b l e in
terms of beliefs and goals, namely REQUEST
and INFORM.
We have
been taking these to
be
prototypical
members
of
Searle's
"directive"
and "representative"
classes
(Searle (1976)).
We represent
questions
as
R E Q U E S T s to
INFORM.
These acts
are
interesting for they have
a wide range of
syntactic realizations, and
account for a
large p r o p o r t i o n of everyday utterances.
Section
4
discusses
the
relation
b e t w e e n the o p e r a t o r s of section 3 and the
linguistic
sentences
which
can
realize
them.
We
c o n c e n t r a t e on
the p r o b l e m
of
identifying
illocutionary
force,
in
particular
on
indirect speech
acts.
A
useful
c o n s e q u e n c e of
the
illocutionary
force
identification process
is that
it
p r o v i d e s a natural way
to u n d e r s t a n d some
elliptical
utterances,
and
utterances
whose purpose
is to
acknowledge, c o r r e c t
or
clarify
interpretations
of
previous
utterances.
(3) We
have limited
o u r s e l v e s so
far to
the
study
of
so-called
task-oriented
dialogues
which
we
interpret
to
be
conversations
between
two
agents
c o o p e r a t i n g in the a c h i e v e m e n t of a single
high-level goal.
These d i a l o g u e s
do not
allow changes
in the
topic of
discourse
but
still
display
a
wide
range
of
linguistic behaviour.
A critical part
of c o m m u n i c a t i o n
is
the
process
by
which
a
speaker
can
c o n s t r u c t d e s c r i p t i o n s of o b j e c t s involved
in his plans
such
that
the hearer
can
identify the
intended referent.
Why can
someone asking "Where's
the s c r e w d r i v e r ? "
be answered
with "In the drawer
with the
hammer" if
it is
assumed he
knows where
the hammer is, but maybe
by "In the third
drawer from the left"
if he doesn't.
How
accurate
must
descriptive
phrases
be?
Section
5 examines
how
the speaker
and
hearer's
m o d e l s of
each other
influence
their
references.
Finally,
section
6
c o n t a i n s some ideas on future research.
Much of our work so far has dealt with
the problem of g e n e r a t i n g plans containing
REQUEST and
INFORM,
as
well
as
nonlinguistic
operators.
Suppose
that
an
agent is a t t e m p t i n g to
achieve some task,
w i t h incomplete k n o w l e d g e of that task and
of the
m e t h o d s to
complete it,
but with
some knowledge of the a b i l i t i e s of another
agent. How can the first agent make use of
the abilities
of the second?
Under what
circumstances
can
the
first
usefully
produce utterances to
transmit or acquire
facts
and
goals?
How
can
he
initiate
action on the part of the second?
Most e x a m p l e s
in the paper
are drawn
from a s i t u a t i o n in which one p a r t i c i p a n t
is
an
information
clerk
at
a
train
station,
whose
objective
is
to
assist
p a s s e n g e r s in boarding and m e e t i n g trains.
The d o m a i n is o b v i o u s l y limited, but still
provides
a natural setting
for
a wide
range of
utterances, both in form
and in
intention.
We view
the plan
related aspects
of
language
generation
and
recognition
as
indissociable, and strongly related to the
process by
which agents c o o p e r a t e
in the
achievement
of goals.
For example,
for
agent2 to reply "It's
closed" to a g e n t l ' s
query
"Where's
the
nearest
service
station?" seems
to require
him to
infer
that
agentl
wants
to m a k e
use
of t h e
service station
which he could not
do if
it were closed.
The
reply "Two
blocks
east" would be seen as m i s l e a d i n g if given
alone, and u n n e c e s s a r y if given along with
"It's closed".
Thus p a r t
of c o o p e r a t i v e
behaviour is the d e t e c t i o n by one a~ent of
obstacles
in the
plans
he b e l i e v e s
the
other agent holds, p o s s i b l y f o l l o w e d by an
attempt to
overcome them.
We
claim that
speakers
expect (and
intend) h e a r e r s
to
operate this
way and
therefore that
any
hearer can a s s u m e that
inferences that he
can draw based on knowledge that is shared
w i t h the
speaker are in fact
intended by
t h e speaker.
These p r o c e s s e s u n d e r ! ~ e our
2.
On m o d e l s
of others
In this
section we
present criteria
that one a g e n t ' s model of another ought to
satisfy.
For
convenience
we
dub
the
agents SELF
and OTHER.
Our
r e s e a r c h has
concentrated
on
modelling
beliefs
and
goals.
We claim that a theory of language
need
not
be
concerned
with
what
is
actually
true
in
the
real
world:
it
should
describe
language
processing
in
terms
of
a person's
beliefs
about
the
world.
Accordingly, SELF's model of O T H E R
should be based on "believe" as described,
for example, in Hintikka(1962)
and not on
"know"
in
its sense
of
"true
belief".
126
Want
Henceforth, all
uses of the
words "know"
and
"knowledge"
are to
be
treated
as
synonyms for "believe"
and "beliefs".
We
have neglected other aspects of a model of
another, such
as focus of
a t t e n t i o n (but
see G r o s z ( 1 9 7 7 ) ) .
Any r e p r e s e n t a t i o n
of O T H E R ' s
goals
(wants) m u s t d i s t i n g u i s h
such information
from:
O T H E R ' S beliefs, SELF's beliefs and
goals, and (recursively)
from the o t h e r ' s
model of someone else's beliefs and goals.
The
representation
for
WANT
must
also
allow for d i f f e r e n t scopes of quantifiers.
For example, it should d i s t i n g u i s h between
the
readings of
"John
wants
to take
a
train" as "There is a specific train which
John wants to
take" or as "John
wants to
take any train".
Finally
it should allow
a r b i t r a r y embeddings with
BELIEVE.
Wants
of
beliefs (as
in "SELF
wants OTHER
to
believe P") become the reasons for telling
P to OTHER, while
beliefs of wants (e.g.,
SELF B e l i e v e s
SELF wants
P) will
be the
way to represent SELF's goals P.
Belief
Clearly,
SELF ought
to
be able
to
distinguish
his beliefs
about the
world
from
what
he
believes
other
believes.
SELF
ought
to have
the
possibility
of
believing
a proposition
P, of
believing
not-P,
or
of
being
ignorant
of
P.
W h a t e v e r his stand on P, he should also be
able to believe that O T H E R can hold any of
these p o s i t i o n s
on P.
N o t i c e that
such
d i s a g r e e m e n t s cannot be represented if the
r e p r e s e n t a t i o n is
based on
"know" as
in
Moore(1977).
Level____~s o f E m b e d d i n g
SELF's belief r e p r e s e n t a t i o n ought to
allow him to represent the fact that O T H E R
knows whether some p r o p o s i t i o n
P is true,
w i t h o u t SELFIs
having to know which
of P
or -P
he does believe.
S u c h information
can
be represented
as
a disjunction
of
beliefs (e.g.,
O R ( O T H E R BELIEVE
P, O T H E R
BELIEVE
~P)).
Such
disjunctions
are
essential
to
the
planning
of
yes/no
questions.
A natural q u e s t i o n to ask is how many
levels of
belief embedding are
needed by
an
agent capable
of
p a r t i c i p a t i n g in
a
dialogue.
Obviously,
to be able
to deal
with a d i s a g r e e m e n t , SELF needs two levels
(SELF
BELIEVE
and
SELF
BELIEVE
OTHER
BELIEVE ).
If SELF were
to lie to OTHER,
he would have
to be able to
believe some
proposition
P
(i.e. SELF
BELIEVE
(P)),
while
O T H E R believes
that SELF
believes
not
P (i.e.
SELF
BELIEVE OTHER
BELIEVE
SELF
BELIEVE (~P)),
and
hence he
would
need at least three levels.
Finally, a belief r e p r e s e n t a t i o n must
distinguish
between situations
like
the
following:
I. OTHER
from gate
2.
OTHER
departure
3. O T H E R
the train
We show
in Cohen
(1978) how
one can
represent,
in
a
finite
fashion,
the
u n b o u n d e d number of beliefs created by any
communication
act
or
by
face-to-face
situations.
The
finite
representation,
which employs
a circular
data structure,
formalizes
the concept
of mutual
belief
(cf.
Schiffer
(1972)).
Typically,
all
these levels
of belief
embedding can
be
represented
in
three
levels,
but
theoretically,
any
finite
number
are
possible.
believes that
the train
leaves
8.
believes
that the
train has
a
gate.
knows what the d e p a r t u r e gate for
is.
Case 1 can be represented by a p r o p o s i t i o n
that c o n t a i n s no variables.
Case 2 can be
represented by
a belief
of a quantified
p r o p o s i t i o n -- i.e.,
OTHER
BELIEVE
x (the y
(
~
GATE(TRAIN,y)
= x))
However,
case
3
is
represented
quantified belief namely,
x OTHER BELIEVE
(the y
:
GATE(TRAIN,y)
by
3.
U§in@ a
What to Say
a
Model
of the Other
to Decide
As a n aid
in evaluating speech
act
definitions,
we
have
constructed
a
computer
program,
OSCAR,
that
plans
a
range
of speech
acts.
The
goal of
the
program
is
to c h a r a c t e r i z e
a speaker's
capacity
to
issue
speech
acts
by
predicting, for specified
situations, all
and only
those speech acts that
would be
a p p r o p r i a t e l y issued by a person under the
circumstances.
In
this section,
we will
make r e f e r e n c e to p r o t o t y p i c a l speakers by
way of
the OSCAR program, and
to hearers
by way of the p r o g r a m ' s user.
= x)
The
formal
semantics such
beliefs
have
been
problematic
for
philosophers
(cf.
Quine
(1956) and
Hintikka (1962)).
Our
approach
to them
is d i s c u s s e d in Cohen
(1978).
In
Section 3,
we
discuss
how
quantified
beliefs
are
used
during
planning,
and how
they
can be
acquired
during conversation.
Specifially,
-
a
127
the program
is able
to:
Plan R E Q U E S T speech
acts, for instance
speech act
that could
be realized
by
"Please
to get
action.
open
the door", w h e n its
goal is
the user to
w a n t to
p e r f o r m some
Suppose,
for example,
that O S C A R
is
outside a
room w h o s e
door is
c l o s e d and
that it b e l i e v e s that
the user is inside.
When
planning
to m o v e
itself
into
the
room, it m i g h t R E Q U E S T
that the user open
the
door.
However,
it w o u l d only
plan
this speech
act if it
believed
that the
user did not a l r e a d y w a n t to open the door
and if it b e l i e v e d
(and b e l i e v e d the user
believed)
that
the
preconditions
to
o p e n i n g the
door held.
If that
were not
so, O S C A R could plan
a d d i t i o n a l I N F O R M or
R E Q U E S T speech acts.
For example, a s s u m e
that to open a door
one needs to have the
key and
OSCAR believes
the user
doesn't
know w h e r e
it is.
Then OSCAR
could plan
"Please open the
door. The key is
in the
closet".
OSCAR
thus
employs
its
user
m o d e l in
telling him w h a t it
b e l i e v e s he
needs to know.
Plan
I N F O R M speech
acts, such
as one
that
could be
realized by
"The door
is
locked", w h e n its goal is
to g e t the user
to b e l i e v e some p r o p o s i t i o n .
-
- Combine
the above to
speech acts in one
plan,
act m a y e s t a b l i s h b e l i e f s
can then
be e m p l o y e d
in
another speech act.
produce multiple
w h e r e one speech
of the user that
the
p l a n n i n g of
Plan q u e s t i o n s
as
r e q u e s t s that
the
user inform,
w h e n its goal is
to b e l i e v e
s o m e t h i n g and
w h e n it
b e l i e v e s that
the
user knows the answer.
P l a n speech
acts i n c o r p o r a t i n g
third
parties, as in "Ask T o m
to tell you w h e r e
the key is and then tell me."
Mediating
Acts
and P e r l o c u t i o n a r y
Effects
The
e f f e c t s of
INFORM (and
REQUEST)
are
modelled
so
that
the
bearer's
b e l i e v i n g P (or w a n t i n g to
do ACT) is not
e s s e n t i a l to the
s u c c e s s f u l c o m p l e t i o n of
the
speech
act.
Speakers,
we
claim,
cannot
i n f l u e n c e their
hearers'
beliefs
and
goals
directly.
Thus,
the
p e r l o c u t i o n a r y e f f e c t s of a s p e e c h act are
not
part of
that
act's d e f i n i t i o n .
We
propose,
then,
as
a
principle
of
c o m m u n i c a t i o n that a
s p e a k e r ' s p u r p o s e in
s i n c e r e c o m m u n i c a t i o n is to p r o d u c e in the
hearer
an a c c u r a t e
model
of his
mental
state.
To i l l u s t r a t e
the p l a n n i n g
of speech
acts,
consider
first
the
following
simplified
definitions
of
REQUEST
and
INFORM as S T R I P S - l i k e o p e r a t o r s (cf. F i k e s
and N i l s s o n
(1971)).
Let
SP d e n o t e
the
speaker, H
the hearer, ACT
some action,
and PROP
some p r o p o s i t i o n .
Due to space
limitations,
the
intuitive
English
m e a n i n g s of the formal
terms a p p e a r i n g in
these d e f i n i t i o n s will h a v e
to s u f f i c e as
explanation.
REQUEST(SP,H,ACT)
preconditions:
SP B E L I E V E H C A N D O A C T
SP B E L I E V E H B E L I E V E H C A N D O A C T
SP B E L I E V E SP W A N T TO R E Q U E S T
effects:
H B E L I E V E SP B E L I E V E SP W A N T H TO ACT
To b r i d g e
the g a p b e t w e e n
the s p e e c h
acts
and
their
intended
perlocutionary
effects,
we p o s i t
m e d i a t i n g acts,
named
C O N V I N C E and
DECIDE, w h i c h model
w h a t it
takes to g e t s o m e o n e
to b e l i e v e s o m e t h i n g
or
want
to do
something.
Our
current
analysis
of
these
mediating
acts
trivializes
the p r o c e s s e s
that they
are
intended
to m o d e l
by
p r o p o s i n g that
to
convince
someone
of
something,
for
example, one need o n l y
g e t that p e r s o n to
know that one b e l i e v e s it.
INFORM(SP,H,PROP)
preconditions:
SP B E L I E V E PROP
SP B E L I E V E SP W A N T TO I N F O R M
effects:
H B E L I E V E SP B E L I E V E PROP
The
program
uses
a
simplistic
backward-chaining
algorithm
that
plans
actions when
their e f f e c t s are
w a n t e d as
subgoals
that
are
not
believed
to
be
true.
It is the
testing of p r e c o n d i t i o n s
of
the
newly
planned
action
before
creating new
subgoals
that e x e r c i s e s the
program's
model of
its
user.
We
shall
b r i e f l y sketch h o w to plan a REQUEST.
Using
Quantified
Questions
E v e r y a c t i o n has "want p r e c o n d i t i o n s " ,
which specify
that b e f o r e
an agent
does
that action, he m u s t w a n t to do it.
OSCAR
plans
REQUEST
speech
acts
to
achieve
precisely
this
precondition
of
actions
that
it
wants
the
user
to
perform.
Similarly,
the
goal
of
the
user's
believing
some p r o p o s i t i o n
PROP
becomes
O S C A R ' S r e a s o n for p l a n n i n g
to INFORM h i m
of PROP.
W h e n such
a q u a n t i f i e d b e l i e f is
a goal,
it leads O S C A R to plan the q u e s t i o n "Where
is the
key?" (i.e.,
R E Q U E S T ( O S C A R , USER,
INFORM(USER,
OSCAR,
the
y
LOC(KEY,y))).
In c r e a t i n g
this q u e s t i o n ,
OSCAR
first
plans a CONVINCE
and
then
p l a n s the u s e r ' s I N F O R M
s p e e c h act, w h i c h
it then tries to get him to p e r f o r m by way
of r e q u e s t i n g .
Beliefs
Notice
that
the
O S C A R ' s g e t t i n g the key
it is -- is of the form:
x
128
--
Planning
precondition
to
-- knowing w h e r e
OSCAR BELIEVE
(the y : LOC(KEY,y)
= x)
The
above
definition
of
INFORM
is
inadequate for dealing with the q u a n t i f i e d
beliefs
that arise
in m o d e l l i n g
someone
else.
This INFORM
should
be viewed
as
that version
of the
speech act
that the
planning
agent
(e.g., OSCAR)
plans
for
itself to
perform.
A different view
of
INFORM, say I N F O R M - B Y - O T H E R ,
is n e c e s s a r y
to represent
acts of informing
by agents
other
than the
speaker.
The
difference
between the
two INFORMs
is that
for the
first, the planner knows
what he wants to
say, but
he o b v i o u s l y does not
have such
knowledge
of the
content
of the
second
act.
The p r e c o n d i t i o n for this new act
quantified speaker-belief:
x
USER BELIEVE
(the y
:
LOC(KEY,y)
the speaker intends to perform, but' as is
well
known,
utterances
which
taken
l i t e r a l l y would indicate one i l l o c u t i o n a r y
force
can be
used
to indicate
another.
Thus "Can
you close the
door?" can
be a
request as well as a question.
These socalled indirect
speech acts are
the acid
test of a theory of speech acts.
We claim
that
a
plan-based
theory
gives
some
insight into this phenomenon.
Searle(1975)
c o r r e c t l y suggests
that
"In cases where
these sentences < i n d i r e c t
forms
of
requests>
are
uttered
as
requests,
they still
have their
literal
m e a n i n g and are uttered with and as having
that literal meaning".
How
then can they
also have their indirect m e a n i n g ?
is a
Our answer relies in part on the fact
that
an
agent
participating
in
a
cooperative dialogue
m u s t have
processes
to:
= x)
where the user is to
be the speaker.
For
the system to plan
an I N F O R M - B Y - O T H E R act
for
the user,
it m u s t
believe that
the
user knows where
the key is, but
it does
not
have
to
know
that
location!
Similarly, the effects
of the
INFORM-BYO T H E R act is also
a quantified belief, as
in
x O S C A R BELIEVE
USER BELIEVE
(the y
.~
LOC(KEY,y)
(I)
Achieve
goals
based
on
what
he
believes.
(2)
Adopt goals
of other
agents as
his
own.
(3) Infer goals of other agents.
(4)
Predict
future
behaviour
of
other
agents.
These p r o c e s s e s would be n e c e s s a r y even if
all speech
acts were l i t e r a l to
account
for e x c h a n g e s where the response indicates
a knowledge
of the
speaker's plan.
For
example
= x)
Thus, O S C A R plans this I N F O R M - B Y - O T H E R act
of
the key's
location in order to
know
where the user thinks the key is.
Passenger: "When does the next train to
M o n t r e a l leave?"
Clerk
: "At 6:15 at Gate 7"
or
Clerk
- "There
won't
be one
until
tomorrow."
Such
information
has
been
lacking
from
all other
formulations
of ASK
(or
INFORM)
that
we
have
seen
in
the
literature (e.g.,
Schank (1975),
Mann et
al. (1976), Searle
(1969)).
Cohen (1978)
p r e s e n t s one approach to defining this new
view
of
INFORM,
and
its
associated
m e d i a t i n g act CONVINCE.
4.
R e c o g n i z i n @ Speech
Speakers
expect
hearers
to
be
e x e c u t i n g these p r o c e s s e s
and they expect
hearers to
know this.
Inferences
that a
hearer
can
draw
by
executing
these
p r o c e s s e s based
on information
he thinks
the speaker
b e l i e v e s can be taken
by the
hearer
to
be intended
by
the
speaker.
This
accounts for
many
of the
standard
e x a m p l e s of
indirect speech
acts such as
"Can you
close the door?" and
"It's cold
here".
For
instance, even if
"It's cold
here"
is
intended
literally
and
is
r e c o g n i z e d as such, the helpful hearer may
still close the window.
When the sentence
is
uttered
as
a
request,
the
speaker
intends
the
hearer
to
recognize
the
s p e a k e r ' s intention that the hearer should
p e r f o r m the helpful behaviour.
Acts
In the
p r e v i o u s section
we d i s c u s s e d
the
structure
of
plans
that
include
instances
of
the o p e r a t o r s
REQUEST
and
INFORM
without
explaining
the
relation
between
these speech
acts and
sentences
used
to
perform
them.
This
section
sketches our first steps in exploring this
relation.
We
have
been
particularly
c o n c e r n e d with the
p r o b l e m of recognizing
illocutionary
force
and
propositional
content of
the utterances
of a
speaker.
Detailed
algorithms
which
handle
the
examples given
in this section
have been
designed
by
J.
Allen
and
are
being
implemented by
him.
Further
d e t a i l s can
be found in (Allen
and P e r r a u l t 1978) and
A l l e n ' s forthcoming Ph.D. d i s s e r t a t i o n .
If
indirect
speech acts
are
to
be
e x p l a i n e d in terms
of inferences speakers
can expect
of hearers,
then a
theory of
speech acts
m u s t concern itself
with how
such
inferences
are
controlled.
Some
h e u r i s t i c s are p a r t i c u l a r l y helpful.
If a
chain of
inference by the hearer
has the
speaker planning
an action
whose effects
Certain
syntactic
clues
in
an
utterance s u c h as its mood
and the use of
explicit performatives
indicate what
act
129
are true
before the
action is executed,
then the chain
is likely to be
wrong, or
else
must
be
continued
further.
This
accounts for "Can you pass the salt?" as a
request for the salt, not a q u e s t i o n about
salt-passing
prowess.
As
Searle(1975)
points
out,
a
crucial
part
of
understanding
indirect
speech
acts
is
being able to recognize
that they are not
to be interpreted literally.
achieve
a goal which
would
allow
deduction
to
continue.
Consider
following example.
Passenger
Clerk
Passenger
Clerk
:
:
:
:
plan
the
When is the Windsor train?
The train to W i n d s o r ?
Yes.
3:15.
After
the
first sentence
the
clerk
cannot
distinguish
between
the
e x p e c t a t i o n s "Passenger travel by train to
Windsor" and
"Passenger meets
train from
W i n d s o r " , so
he sets up
a goal
: (clerk
believes
passenger
wants to
travel)
or
(clerk
believes passenger
wants to
meet
train).
The
planning
for
this
goal
p r o d u c e s a plan that involves
asking the
passenger
if
he
wants
one
of
the
alternatives,
and
receiving
back
the
answer.
The
execution
of
this
plan
p r o d u c e s the clerk response
"The train to
Windsor?"
and
recognizes
the
response
"Yes".
Once
the
passenger's
goal
is
known, the clerk can c o n t i n u e the o r i g i n a l
deduction
process
with
the
"travel
to
Windsor" a l t e r n a t i v e favoured.
This plan
is accepted
and the
clerk p r o d u c e s
the
response "3:15"
to overcome
the o b s t a c l e
"passenger knows d e p a r t u r e time".
A second h e u r i s t i c is
that a chain of
inference that
leads to
an action
whose
p r e c o n d i t i o n s are
known to be
not easily
a c h i e v a b l e is likely to be wrong.
Inferencing
can
also
be
controlled
through the use of
e x p e c t a t i o n s about the
s p e a k e r ' s goals.
P r i o r i t y can be given to
inferences which relate an o b s e r v e d speech
act
to
an expected
goal.
Expectations
enable
inferencing
to work
top-down
as
well as bottom-up.
The use of expected goals to guide the
inferencing
has
another
advantage:
it
allows
for
the
recognition
of
illocutionary
force
in
elliptical
utterances
such
as "The
3:15
train
to
Windsor?",
without
requiring
that
the
syntactic
and
semantic
analysis
"reconstitute"
a
complete
semantic
representation
such
as "Where
does
the
3:15
train
to W i n d s o r
leave?".
For
example,
let
the
clerk
assume
that
passengers
want to
either m e e t
incoming
trains or board d e p a r t i n g
ones.
Then the
utterance "The 3:15 train
to W i n d s o r ? " is
first
interpreted as
a REQUEST about
a
train
to Windsor
with
3:15
as
either
arrival or d e p a r t u r e time.
Only d e p a r t i n g
trains
have d e s t i n a t i o n s
different
from
T o r o n t o and
this leads to
believing that
the passenger wants to
board a 3:15 train
to
Windsor.
Attempting
to
identify
o b s t a c l e s in the p a s s e n g e r ' s plan leads to
finding that the p a s s e n g e r
knows the time
but p r o b a b l y
not the place
of departure.
Finally,
overcoming
the
obstacle
then
leads to an INFORM like "Gate 8".
5.
Reference
and
the Model
of the Other
We have shown that q u a n t i f i e d beliefs
are needed
in d e c i d i n g
to ask
someone a
question.
They
are
also
involved,
we
claim, in
the r e p r e s e n t a t i o n
of singular
definite
noun
phrases
and
hence
any
natural
language system
will need
them.
According to our analysis, a hearer should
represent
the
referring
phrase
in
a
s p e a k e r ' s s t a t e m e n t "The pilot
of TWA 510
is drunk" by:
x S P E A K E R BELIEVE
(the y
:
PILOT(y,TWA510)
DRUNK (x))
= x &
This is the reading w h e r e b y the speaker is
b e l i e v e d to "know who the pilot of TW~ 510
is"
(at
least p a r t i a l l y
accounting
for
Donnellan's
(1966) referential
reading).
This is to be
c o n t r a s t e d with the reading
of whoever is p i l o t i n g that plane is drunk
(Donnellan's
attributive
noun
phrases).
In
this
latter
case,
the
existential
q u a n t i f i e r would
be inside
the scope
of
the belief.
Our analysis of
elliptical u t t e r a n c e s
raises
two
questions.
First,
what
information does
the i l l o c u t i o n a r y
force
r e c o g n i t i o n m o d u l e expect
from the syntax
and semantics? Our approach
here has been
to require from the syntax and semantics a
h y p o t h e s i s about the literal i l l o c u t i o n a r y
force
and
a
predicate
calculus-like
representation
of
the
propositional
content, but where u n d e t e r m i n e d p r e d i c a t e s
and objects could be
replaced by p a t t e r n s
on
which
certain
restrictions
can
be
imposed.
As part of
the plan inferencing
process
these
patterns
become
further
specified.
These
existential presuppositions
of
d e f i n i t e referential noun p h r a s e s give one
important
way
for
hearers
to
acquire
quantified s p e a k e r - b e l i e f s .
Such beliefs,
we have seen, can be used as the basis for
p l a n n i n g further c l a r i f i c a t i o n questions.
The
second q u e s t i o n
is: w h a t
should
the
hearer
do
if more
than
one
path
between
the
observed u t t e r a n c e
and
the
e x p e c t a t i o n s is p o s s i b l e ?
He may suspend
plan
deduction
and
start
planning
to
We
agree
with S t r a w s o n
(1950)
(and
many
others)
that
hearers
understand
referring
phrases
based
on
what
they
believe
speakers
intend
to
refer
to.
130
Undoubtedly,
a hearer
will understand
a
s p e a k e r ' s (reference) intentions
by using
a
model
of
that
speaker's
beliefs.
Speakers,
of
course,
know
of
these
interpretation
s t r a t e g i e s and
thus
plan
their
referring
phrases
to
take
the
a p p r o p r i a t e referent
within the
hearer's
model
of
them.
A speaker
cannot
use
private
descriptions,
nor
descriptions
that
he
thinks
the
hearer
thinks
are
private, for c o m m u n i c a t i o n .
handle p r o m i s e s w i t h o u t first d e a l i n g with
obligations,
or
warnings
without
the
n o t i o n s of danger
and u n d e s i r a b i l i t y ?
We
are c u r r e n t l y c o n s i d e r i n g
an e x t e n s i o n of
the
approach
to
understanding
stories
which report simple dialogue.
Much
remains
to
be
done
on
the
r e p r e s e n t a t i o n of the a b i l i t i e s of angther
agent.
A simple setting suggests a number
of problems.
Let one agent H be seated in
a
room
in
front
of
a
table
with
a
c o l l e c t i o n of
blocks.
Let
another agent
S be outside the room but c o m m u n i c a t i n g by
telephone.
If S b e l i e v e s
that there is a
green
block on
the
table
and wants
it
cleared, but knows nothing about any other
blocks except
that H can see
them, then
how can S ask H
to clear the g r e e n block?
The blocks S wants removed are those which
are in fact there,
p e r h a p s those which he
could p e r c e i v e to
be there if he
were in
the room.
The goal
seems to
be of
the
form
For instance,
consider the
following
variant
of
an
example
of
Donnellan's
(1966):
At a party, a woman is holding a
martini
glass
which
Jones
believes
contains water, but of which he is c e r t a i n
everyone
else believes
(and b e l i e v e s
he
believes) contains a martini.
Jones would
understand that
Smith, via
q u e s t i o n (I),
but not
via q u e s t i o n (2) is
referring to
this woman.
(i) Who
(2) Who
is the woman
is the woman
holding
holding
since Jones
does not believe
about the water in her glass.
the m a r t i n i ?
the water?
Smith
S BELIEVE
x (x on the green block => S WANT
(x removed from green block))
knows
Conversely, if Jones wanted
to refer
to the woman in an u t t e r a n c e intended for
Smith, he
could do so
using (i)
but not
(2) since in the latter
case he would not
think
the
hearer
could
pick
out
his
intended referent.
but our planning
m a c h i n e r y and d e f i n i t i o n
of R E Q U E S T
are inadequate
for g e n e r a t i n g
"I request you to clear the green block".
We
have
not
yet
spent
much
time
investigating
the
process
of
giving
answers to How and Why questions, or to WH
q u e s t i o n s requiring
an event
description
as an answer.
We
c o n j e c t u r e that because
of
the
speech act
approach
answers
to
"What did he say?" should be found in much
the same
way as answers
to "What
did he
do?"
and
that
this
parallelism
should
extend
to other
question
types.
The
natural
e x t e n s i o n of
our a n a l y s i s
would
suggest r e p r e s e n t i n g "How
did AGT achieve
goal G?" as a R E Q U E S T
by the speaker that
the hearer inform
him of a plan
by which
AGT
achieved
G.
We
have
not
yet
investigated
the
repercussions
of
this
e x t e n s i o n on the r e p r e s e n t a t i o n language.
Thus it appears that
for a speaker to
plan
a
successful
singular
definite
referential e x p r e s s i o n
requires that
the
speaker believe the
e x p r e s s i o n he finally
chooses
have the
right
referent in
the
hearer's
model
of
the
speaker.
Our
c o n c e p t of
mutual belief can be
used (as
in Cohen
(1978))
to
ensure
that
the
expression
denotes appropriately
in
all
further
embedded
belief
models.
This
example is p r o b l e m a t i c for any approach to
reference
where
a
communicating
party
assumes
that
its
reality
is
the
only
reality.
Speakers
and
hearers
can
be
"wrong"
or
"ignorant"
and
yet
c o m m u n i c a t i o n can still
be m e a n i n g f u l and
successful.
Finally
consider
the
following
dialogue.
Assume
that
S
is
a
shady
businessman, A his secretary.
A : IRS
6.
Further
is on the phone.
Research
S : I'm not here.
We believe that speech acts p r o v i d e an
e x c e l l e n t way of
e x p l a i n i n g the relations
between utterances in a dialogue, as well
as relating
linguistic to
non-linguistic
activity.
Until we
better understand the
m e c h a n i s m s by which c o n v e r s a n t s change the
topic
and goals
of
the c o n v e r s a t i o n
it
will be d i f f i c u l t to
extend this analysis
beyond e x c h a n g e s
of a few
utterances, in
particular to non-task oriented dialogues.
Fuller j u s t i f i c a t i o n of
our approach also
requires its a p p l i c a t i o n to a much broader
range of speech acts.
Here the p r o b l e m is
mainly
representational:
how
can
we
How
is A to
understand
S's
utterance?
Although
its
propositional
content
is
l i t e r a l l y false,
maybe even
nonsensical,
the u t t e r a n c e ' s intention is unmistakable.
How tolerant does the u n d e r s t a n d i n g system
have to be
to infer its way
to a c o r r e c t
interpretation?
Must "I'm
not here"
be
treated i d i o m a t i c a l l y ?
131
B ibl log r aphy
Allen,
J.F.
and
Perrault,
C.R.,
"Participating
in
Dialogue:
Understanding via Plan Deduction", 2nd
National Conference
of the Canadian
Society
for Studies in Computational
Intelligence, Toronto, July, 1978.
Cohen,
P.R.,
"On Knowing
What
to Say:
Planning Speech Acts", TRII8 Dept.
of
Computer
Science,
University
of
Toronto, 1978.
Donnellan,
K.,
"Reference
and Definite
Description",
The
Philosophical
Review,
vol.
75,
1960,
pp280-304.
Reprinted in Semantics, Steinberg
and
Jacobovits, eds., Cambridge University
Press, 1970.
Fikes,
R. E. and Nilsson, N. J., 1970,
"STRIPS:
A new
approach
to
the
application
of
theorem
proving",
Artificial Intelligence 2, 1970.
Grosz,
B. J., "The Representation and Use
of
Focus
in
Natural
Language
Dialogues", 5IJCAI, 1977.
Hintikka,
K.J.,
Knowled~[e
and Belief,
Cornell University Press, 1962.
Mann, W.C.,
Moore, J.A., Levin, J.A.; "A
Comprehension
Model
for
Human
Dialogue", 5IJCAI, 1977.
Moore,
R.C.;
"Reasoning
about Knowledge
and Action", 5IJCAI, 1977.
Quine,
w.v.,
"Quantifiers
and
Propositional Attitudes", The Journal
of Philosophy 53, (1956), 177-187.
Schiffer,
S., Meaning, Oxford University
Press, 1972.
Schank,
R. and Abelson,
R.,
"Scripts,
Plans and Knowledge", 4IJCAI, 1975.
Searle,
J. R.,
Speech
Acts,
University Press, 1969.
Cambridge
Searle,
J. R.; "Indirect Speech Acts" in
Syntax and Semantics, Vol.
3: Speech
Acts,
Cole and Morgan (eds), Academic
Press, 1975.
Searle,
J.
R.,
"A
Taxonomy
of
Illocutionary
Acts",
Language,
Mind
and Knowledge,
K. Gunderson
(ed.),
University of Minnesota Press, 1976.
Strawson,
1950.
P.
F.,
"On
Referring",
Mind,
132