Você está na página 1de 3

Three Typed Pragmatics for Dialogue Structure Analysis

Hitoshi IIDA*, Takayuki YAMAOKA* and Hidekazu ARITA**


* A'I2R Interpreting Telephony Research Laboratories
Sanpeidani, hmidani, Seika-cho,Souraku-gun, Kyoto 619-02, ,Japan
[{iida/yamaoka}%atr-la.atr.co.jp@uunet. UU.NET]
** Mitsubishi Electoric Corporation
8-1-1 Tsukaguchi-honmaehi,
Amagasaki, tlyougo 661, J a p a n
[arita%sys.crl.melco j unet@uuneL UU .NETI

Abstract Recently, a plan-based dialogue


An experimental system for dialogue understanding approach was developed using a
structure analysis based on a new type plan kind of pragmatics metaplan: a 'discourse plan'
recognition model for spoken dialogues has been incorporated with domain knowledge, 'domain
implemented. This model is realized by using plan' [Litnmn871. This approach was based on
four typed plans which are categorized into three Allen & Perrault's plan recognition s t r a t e g i e s
kinds of universal pragmatics and a ldnd of task- [AllenS0], [PerraultS0], [Allen871. By way of
dependent knowledge related to common action contrast, a pragmatic-based approach was
h i e r a r c h i e s . The e x p e r i m e n t a l s y s t e m is proposed in order to understand intersentential
c h a r a c t e r i z e d by h i g h e r m o d u l a r i t y a n d elliptical fragments [Carberry891. She claimed
computational efficiency t h r o u g h d e f i n i n g a L i t n m n ' s s t r a t e g i e s c o u l d not r e c o g n i z e a
hierarchical usage order between these surprise o r s doubt conveyed by an elliptical
knowledges. The system can grasp a dialogue fragment, for example "$10,000 ":",. which is not a
structure making it possible to solve problems complete sentence including postulated speech
related to spoken dialogue interpretation. acts 'request' or 'inform'. In addition, she also
claimed that metaplans representing
1. INTRODUCTION c o m m u n i c a t i v e goals should be d e a l t w i t h
An efficient, and s m o o t h c o m m u n i c a t i o n adequately. She newly introduced ' d i s c o u r s e
be'~ween humans is generally realized in spoken e x p e c t a t i o n ' r u l e s for g r a s p i n g i n t e r a c t i o n
dialogues. This fact is m a i n l y s u p p o r t e d by b e t w e e n an i n f o r m a t i o n - s e e k e r a n d an
v a r i o u s ellipsis exp:°essions c o n c e r n i n g old information-provider and 'discourse goal' rules
information, the dialogue participants, zero- for i d e n t i f y i n g an i n f o r m a t i o n - s e e k e r ' s
pronouns - especially in dapanese, substitutional conversational or c o m m u n i c a t i v e goa[*(l).
verbs and so on. A s s result, each utterance is These rules are heuristics for interpreting an
fragmental. A sequence of t h e s e u t t e r a n c e s elliptical fragment which explicitly indicates no
generally construct a whole dialogue step by linguistic clues to interpreting speech act.s. On
step. ['ragmental utterance comprehension by a the other hand, in order to make the general
hearer can be achieved using knowledge of the understanding mechanism clear, a surprise or a
dialogue sitamtion, context intbrmation, domain doubt fragment such as mentioned above, must
d e p e n d e n t knowledge, especially the donmin be understood from recognizing an i r r e g u l a r
dependent action hierarchy[Litman871, meaning from the view of gaps between common
universal pragmatics concerning how to advance sense (a certain prerequisite eondition) and the
a dialogue, maintain dialogue cooperation d e c l a r a t i o n . F u r t h e r m o r e , s t u d y i n g how to
between dialogue participants, etc., and express a query, an answer, a confirmation, a
language specific pragmatics [ Levinson831. Such surprise or a doubt in a context and a dialogue
ellii;sis-resolution-is one of the main problems in situation is necessary.
discourse understanding. Several approaches In c o n s i d e r a t i o n of m a k i n g a d i a l o g u e
were proposed and implemented in [fiendrix781, u n d e r s t a n d i n g m o d e l i n c l u d i n g s u c h an
[Shimazu79], [Carbonel1831, etc. These u n d e r s t a n d i n g process, a m e t h o d to h a n d l e
researchers used various heuristic rules and did pragmaties and utterance or dialogue must be
not make a clear distinction between domain s t u d i e d . T h i s l?aj)er s h o w s t h r e e t y p e d
~pecific k n o w l e d g e and p r a g m a t i c s . As a p r a g m a t i e s use(t 1or c o o p e r a t i v e d i a l o g u e
consequence, the user interface became development, as well as a dialogue s t r u c t u r e
inflexible because the system basically handled analysis and understandinfi model using a plan
expected utterance patterns. recognition approach. 'lhe p r a g m a t l c s are
described by a 'dialogue plan', 'communication
plan' and 'interaction plan'.
*(1) :A discourse goal is characterized by a discourse
expectation which dialogue participants mutually believe 2. THREE TYPED PRAGMATICS
as an expectation. Therefore, a discourse goal is very vague Three types of universal p r a g m a t i e s (el.
because a expectation depends on various attentions or LevinsonS3) can be classified and described by
empathies and the participant's knowledge, the following plans: Interaction-Plan - a plan
*(2) :Allowingembedded turn-takings. basically e h a r a e t e r i z e d by a d i a l o g u e t u r n -
*(3):A communicative act is basically defined as an taking*(2) w h i c h d e s c r i b e s a s e q u e n c e of
abstract action, one that effects the bearer's thinking or communicative acts*(3), Communication-Plan -
deci,qiommaking and which can be described by a plan a plan which d e t e r m i n e s how to e x e c u t e or
schema (cf.Cohen84).

370
achieve an utterance goal or dialogue goals, and domain..specific, knowledge is related to the
Dialogued:qan : a plan for e s t a b l i s h i n g a action and objects, e s p e c i a l l y the a c t i o n
d i a l o g u e c o n s t r u c t i o n , e.g. a c o o p e r a t i v e hierarchy which is used to grasp the utterance
dialogue*(4).
For example, in order to achieve the goal of ~oal, and three kinds of pragmatics between
unmns involved in a spoken dialogue are used
registering for the conference tim following to grasp the dialogue development.
sequer, ce must usually be performed (Domain-
Plan}: obtain a registration form, fill out the 3A Communicative Acts
form and return it to the secretariat. In such a First, in order to recognize the speaker's
telephone dialo}'ue, if something is needed to plan, it is necessary to recognize the turn-taking
execute the gore action, a request to send it can patterns. Communicative acts [Cohen84] are
be m a d e , or it will be o f f e r e d to y o u i n t r o d u c e d . In F i g . l , an e x a m p l e of
( C o : m m u n i c a t i o n - t ) l a n ) . To c o m p l e t e the communicative acts in a cooperative task-
cooperative infbrnmtion-seeking, the hearer will oriented dialogue, e.g. 'queries and explanations
resl)ond to the speaker's request*(5) (Interaction- regarding r e g i s t r a t i o n for an i n t e r n a t i o n a l
Plan). Beiorc b u i l d i n g a whole d i a l o g u e conference' is shown.
structure, the speaker should utter the opening A communicative act in the demand class
section of the dialogue, e s p e c i a l l y on the and a corresponding act in the response class
telephone. Furthermore when the dialogue is make a turn-taking pair. This is recognized by
finished, the s p e a k e r should wide up the the interaction plan. A communicative act is a
dialogue (Dialogue-Plan). decomposition element of an interaction plan.
Each plan is described in terms of a schema
formulation (plan-schema). Denmnd Class Response Class
A plan-schema has various slots to describe both
an action's inner properties, e.g. HEADER and Ask-value Inform-value
P I { L I ) I C A I E & C A S E S , and r e l a t i o n s t t l p s
41 r~ ,l • • "Whet, is the deadline? . . . . The deadline for the
b e t w e e n the a c t i o n a n d p r e r e q u i s i t e paper is,J une 3."
Confirm-value . Affirmative
states/actions, effects, etc., e.g. PREREQUISITE, "The deadline is Jtme 3, isn't it? . . . . Yes, it is."
DECOMPOSITIONS, CONSTRAINTS, Negative
EFFECTS. "No, it isn't."
A definite hierarchical order among these Request-action Accept
plans is available as follows; "Will you send me the lbrm? . . . . OK"
Interaction-Plan > Communication-Plan > Reject
Domain-Plan > Dialogue-Plan. "I'm afraid I can't."
Give..offer Accept-oiler
3. DIALOGUE ANALYSIS "Shall I send you the tbrm?" "Yes, please"
There are several linguistic p h e n o m e n a Reject.offer
which are hard to interpret, such as ellipses, "No, thank you."
referring pronouns and substitutional Eig.1 At) Example oi' Commurticative Acts
e x p r e s s i o n s . Both i n f o r m a t i o n f r o m the
established context and expectations from the 3.2 Inference Mechanism
c u r r e n t dialogue situation are r e q u i r e d to 'An utterance meaning is represented by
resolve such problems. In order to get these illocutionary speech act types and propositional
information a dialogue structure which indicates contents obtained from a t l e a d - D r i v e n
the goal hierarchy of utterances in a dialogue unification-based active chart parser*(6) for
must be constructed. A dialogue analysis is J a p a n e s e dialogues. The parsing r e s u l t is
required and it is necessary to determine ; described by a feature structure and tim system
(1) how to infer each goal of an utterance input is modified into a communicative act with
within a dialogue, propositional contents. These consist of a certain
(2) how to make clear the relationships predicate, an ' u t t e r a n c e type', and some
between goals within the dialogue. variables, in particular, 'speaker', 'hearer', and
For the first problem, a plan recognition 'topic'. 'Topic' is, on a surface level, an NP
inference method is adequate for identifying an marked with the Japanese special particle, 'wa',
utterance intention because the intention can be or the compound particle 'nituite'.
inferred by recognizing the speaker's plan by The plan recognizer (1) assumes a goal. (2) If
chaining communicative acts regarded as speech a particular goal cannot be found, then stop else
acts in a specific domain [Allen80), [Perrault80], goto next (3). (3) Infer a chaining path from an
[Litman 87]. For the second problem, the input to the goal. If success, stop. Else return to
the first process (1) in order to try to find the next
candidate. The chaining process between plans
*(4) :A dialogue global construction usually has an opening g e n e r a l l y finds a candidate plan from the
section and closing section. I lere, such a linguistic c u r r e n t state (IIEADER) to an action list
phenotmnon is regarded as language-universal. represented in I)ECOMPOSITtON. Ilowever if
*(5) :On the other hand, in order to complete the cooperative this fails, the chaining will be continued in
interaction, when the speaker imparts information, the accordance with PREREQUISITE and EFFECT.
hearer will confirm what the speaker has said according to In o r d e r to m a n a g e the c u r r e n t
the speaker's belief in the hearer's intention. understanding state, the system uses two slacks.
*(6) An active chart parser has been developed for UN1)ERSTANDING-IAST stores completed
Japanese dialogue analysis on a unification based grammar, plans as the current understanding state, and
which is based on IIPSG and J P S G [Pollard88), [Gunji87], GOAL-IAST m a i n t a i n s i n c o m p l e t e p l a n s
[Kogure89]. Furthermore, many discourse entities can be regarded as possibilities and expectations for
identified by using NP Identification Method [Nogaito88].

-2- 371
future goals . An overview of a dialogue colleagues at ATR Interpreting Telephony
s t r u c t u r e c o n s t r u c t i o n process is shown in Fig.2. R e s e a r c h L a b o r a t o r i e s for t h e i r e n c o u r a g e m e n t
and t h o u g h t - p r o v o k i n g discussions.
4. A N E X P E R I M E N T A L SYSTEM
This a n a l y s i s model is realized by using lout" REFERENCES
typed p l a n s w h i c h a r e c a t e g o r i z e d into Ulree [Allen 801 Allen, J. F. and PerraulL, C. R. : Analyzing
types of u n i v e r s a l p r a g m a t i c s and a type o£ task- Intention in Utterances,
d e p e n d e n t knowledge r e l a t e d to c o m m o n action Artificial Intelligence, Vol.15, pp143-178(1980L
hierarchies. The s y s t e m has been i m p l e m e n t e d [Allen87] Allen, J. F. : N a t u r a l Language
Understanding, The BeujaminlCummings Publishing Co.
in Symbolics C o m m o n Lisp. A dialogue s t r u c t u r e (1987).
is r e p r e s e n t e d by both c o m p l e t e d p l a n s a n d [Carberry 891 Carberry, S. : A Pragmatics-Based Approach
i n c o m p l e t e p l a n s s t o r e d in t h e two s t a c k s . to Ellipsis Resolution,
T h e r e f o r e , the s y s t e m can u n d e r s t a n d dialogue Computational Linguistics, Vol.15, No.2, pp75-96(1989).
m e a n i n g s and c a n offer a d i a l o g u e s t r u c t u r e [Carbonel183]Carbonell, J. : Discourse Pragmatics and
using the contents of both stacks. F o u r m o d e l Ellipsis Resolution in Task-Oriented Natural Language
dialogues r e g a r d i n g an i n t e r n a t i o n a l c o n f e r e n c e Interfaces, 21st Annual Meeting of the ACL,pp164-
r e g i s t r a t i o n t a k e n from slightly modified inter- 168(1983).
k e y b o a r d d i a l o g u e s in J a p a n e s e h a v e b e e n [Cohen 84] Cohen, P. R. : The Pragmatics or Referring and
Modality of Communication, Computational Linguistics,
applied. For example, the s y s t e m can u n d e r s t a n d Vol.10, No.3, pp97-146(1984L
a J a p a n e s e s u b s t i t u t i o n a l e x p r e s s i o n , e.g. "O- [ttendrix 781 Hendrix, G. G. et al. : Developing a Natural
isog-i k u d a s a i " ( l i t e r a l t r a n s l a t i o n : ' H u r r y Language Interface tx) Complex Data, ACM Trans, Vol.3,
please') which shows no agent, no object and no No.2, pp 105-147(i 978).
verb, because the c u r r e n t topic which is focused [Gunji87] Gunji, T, : Japanese Phrase Structure
on an action in the d o m a i n plan is known in the Grammar, Dordrecht, D. Reidel (1987).
s y s t e m and the o m i t t e d verb (e.g. ' R e t u r n - F o r m ' ) [lida 891 lida, II. eL al. : An Experimental Spoken Natural
can be identified u n d e r the scope d o m i n a t e d by Dialogue Translation System Using a Lexicon-I)riven
the topic. Grammar, European Conference on Speech Communications
and Technology (1989).
5. CONCLUSION [Kogure891 Kogure, K. : Parsing Japanese Spoken
Sentences Based eta HPSG,
L i t m a n & Allen introduced a set of discourse the hit. Workshop on Parsing Technologies (1989).
plans, each one c o r r e s p o n d i n g to a p a r t i c u l a r [Levinson 83] Levinson, S.' C. : Pragmatics, Cambridge
way t h a t an u t t e r a n c e can r e l a t e to a discourse University Press (1983).
topic. T h e y d i s t i n g u i s h discourse plans f r o m a [l,itman871 Litman, D: J. and Allen, J. F. : A Plata
set of d o m a i n p l a n s . T h e d i a l o g u e s t r u c t u r e Recognition Model for Subdialogues in Conversations,
a n a l y s i s model basically follows the above idea Cognitive Science, Vol.i 1, pp163-200 (1987J.
and uses new three typed pragmatics: [Nogaito881 Nogaito, I. and lida, ti. : Noun l'hrase
interaction plan, communication plan and Identification iu Dialogue and its Application, 2nd
dialogue plan. By introducing these plans, the International Conf. on Theoretical and Methodological Issues
in Machine Translation ofNatural I,anguage (1988L
mech~mism for c o n s t r u c t i n g a dialogue s t r u c t u r e [Perrault 801 Perrault, ~. R. and Allen, J. F. : A Plan-Based
b e c o m e- s c l e a r b e c a u s e of the w a y a s u r f a c e Analysis of Indirect Speecia Acts, Computational Linguistics,
u t t e r a n c e is connected with both p r a g m a t i c s and Vol.6, pp166-182(1980).
the domain-specific knowledge, and by r e d u c i n g [Polard891 Pollard, C. and Sag, I. : lnfornaation-Based
the s e a r c h space using a h i e r a r c h i c a l o r d e r of Syntax and Semantics - Vol.l Fundamentals, CSLI Lecture
a p p l y i n g knowledge, c o m p u t a t i o n a l efficiency is Notes, No.13 (I988).
improved. [Shimazu79] Shimazu, A. and lida, If. : Experimental
Japanese l,anguage Question Answering System MSSS78,
Acknowledgement 6th International Joint Conf. on AI, pp803-808 (1979).
T h e a u t h o r s w o u l d l i k e to t h a n k T h e
P r e s i d e n t Dr. A k i r a K u r e m a t s u and our o t h e r
Loi
a_~_931oue structr_Lg]
u
: D ~ a ~ ~ Previous utterances
- - t }~ o-~R'T; MAKE, R E S i ~ , ~ (Interpretation) /
IDECOMPOSITION; [ GET-FOaM J /" spl: I'd like to make a registration for the conference,
FILL-FOR~a Spl: W h a t should I do7 ,

,.roon
sph Please send me

__ G,T-EO~M DIRECTION ..... .°,,,O,


,p~.k.rC ..... ,y~ t.,Oo.T.A~,o.l,,.,,,
,P' 1 ~.
I TopIc form
IOemai.nP!l.anI
J.~ADE.: SE.O-EO~
PREREQUISITE: KNOW(ADORESS&NAME)
J
IEFFECT: HAVE -A-FORM J
T----
(Predlctlon$) If_communication
IL---~EADER:
Plan i
iNTRODUCE-DOMAiN-PLAN [interaction Planl
r" . . . . "I [OECOMPOSlI"ION: [ REQUES]-ACTION*UNI-~ " - - ~ H E A D £ R : REQUEST-ACXION-UNIT]
• . . . . .
~ /i//l//////z///
. .

.... L.IFILL'FO~M!.~ (Predkeon$)[ [ WILL-OO.ACTION,UNIlJ ]|DEEOMPOSl]ION; REOUEST.ACTION ~ - - ~ REQUEST-ACTION ~'~


: ,- ...... , ~- ...... . l ActEd,, --------Jl ~/,,//,-////(~z. ((./..~
.... I R E T U R N - F O R M I .... I ACHIEVE.KNOW | (Predictions) J... r[ACCEPT
-----'*
I "':: .......................
Predl~tln9candidate ::
; {sp2: All ¢lghl. :
........................ .

Fig.2 A n O v e r v i e w of a D i a l o g u e S t r u c t u r e C o n s t r u c t i o n Process

379,
3 -

Você também pode gostar