COMMUNICATIVE TRIAD 
AS A STRUCTURAL ELEMENT OF LANOUAGE INTERACTION 
F.G.Dinenberg 
AI Laboratory, Computer Center 
Siberian Division of the USSR Ac. Sci. 
Novosibirsk 630090, USSR 
ABSTRACT 
Researches on dialogue natural-language inter- 
action with intellectual "human-computer" systems 
are based on models of language "human-to-human" 
interaction, these models representing descriptions 
of communication laws. An aspect of developing lan- 
guage interaction models is an investigation of 
dialogue structure. In the paper a notion of elemen- 
tary communicative triad (SR-triad) is introduced to 
model the "stimulus-reaction" relation between 
utterances in the dialogue. The use of the SR-triad 
apparatus allows us to represent a scheme of any 
dialogue as a triad structure. SR-triad structure 
being inherent both to natural and programming 
language dialogues, SR-system is claimed to be 
necessary while developing dialogue processors. 
1. INTRODUCTION 
In earlier papers devoted to interpersonal 
interaction iFrank,1981; Levinson,1981\] much atten- 
tion is paid to studying the role of speecb act (SA) 
in dialogue structure. Considering SA as a principal 
functional element of language interaction (LI) is 
of special importance for developing natural 
language models which are to be practically applied 
to human-computer natural--language interaction. 
We proceed from the statement that there are 
not solitary performed speech acts in real communi- 
cation, SAs are interconnected and involved in a 
general LI structure \[Wunderlich,1980\]. 
Within the framework of our approach to the 
formal\[ description of dialogue structure a notion of 
SR-triad reflecting the "stimulus-reaction" relation 
between utterances in the dialogue is proposed to 
construct a more adequate representation of dialogue 
structure. Representing any dialogue as a triad 
structure (T-structure) in contrast to previous 
analysis \[Hundsnurscher,1981\] is characterized by 
distinguishing three phases in the interaction of 
the type "stimulus-reaction". First we shall 
consider different roles of speech acts in the 
dialogue T-structure. Then we shall introduce three 
types of relations between SR-triads (intersection, 
imbedding and succession) and discuss dialogues with 
a complicated structure which is represented by 
combinations of these relations and reflects diffe- 
rent strategies to attain some goal. 
2. SR-TRIAD AND A TYPOLOGY OF SAS 
Any dialogue is considered to be a sequence of 
SAs distributed in time which is charaterized by: a) 
a change of roles of the speaker and the hearer 
between communication participants (further desig- 
nated as X and Y) and b) certain relations between 
SAs. 
Each SA is aimed at accomplishing whole complex 
of goals/tasks. The discussion of their content is 
not relevant to the purposes of this paper. So we 
confine our discussion to one goal (g.X) only asso- 
ciated with some SA of the author X (SA.X). The very 
fact of producing an SA which "initiates" g.X raises 
before the addressee Y a communicative task to react 
to this goal, i.e. to close g.X positively or nega- 
tively. Thus between SA.X and the following SAs of Y 
there appear "stimulus- reaction" relations. Between 
an initiation and a closure of g.X there may take 
place intermediate interactions aimed at accomplis- 
hing the communicative task by Y. The intermediate 
SAs of Y are associated with goals subordinated to 
the communicative task, and the intermediate SAs of 
X which are reactions to the corresponding SAs- 
stlmuli of Y are associated with the initiating SA.X 
by the goal g.X and aimed at explanation, support- 
ing, correction of the goal or some other components 
of tile SA.X content. 
2.1. In the capacity of the basic structural 
element of dialogue interaction, SR-trlad is pro-- 
posed which reflects three following phases of the 
interaction these phases being associated with any 
goal initiated in the dialogue. 
The first ~hase is a speech act SA1.X by means 
of which its author X induces his dialogue partner Y 
to realize some communicative goal g.X (initiation 
of g.X). 
The ~#cond phase is represented by a positive 
or negative reaction SA1.Y of Y to this goal (clo- 
sure of g.X). 
The third ~hase is a reaction SA2.X of X to 
SAI.Y which may be positive (acceptance) or negative 
(rejection) as well. Often SA2.X may be absent in 
real dialogues that means an acceptance of Y's reac- 
tion by default (as it follows from the completion 
of the interaction or from the transition to the 
initiation of' a new goal). Irrespective of the 
explicit/implicit form of expression, this phase is 
considered as a regular organizing component of the 
interaction process. 
2.2. Let us enumerate principal types of speech 
acts reflecting the roles of the SA in the dialogue 
T-structure and introduce symbols to designate them 
in schemes. 
(1) An initiating SA is a primary initiation of 
the goal (see the designation in fig. la).*) 
(2) An initiative-reactive SA is an initiation 
of an intermediate goal without trying to close the 
principal goal, for example, the author of the given 
SA has not sufficient information for generating a 
reaction to the principal goal, or the information 
at the author's disposal is considered by the 
author as non-authentic, gives rise to doubt, needs 
correction etc. Such an SA is aimed at some 
component of the content of the previous SA 
(the connection with this component is designated by 
a dotted line (see fig.lb)). 
(3) A reactive SA is a positive or negative 
reaction (to the goal) reflecting the second phase 
of an SR-triad (fig.lc). This reaction may be 
*)Here and further (in schemes) a horizontal line 
represents a goal with which the role of the given 
SA is connected in an SR-triad; a line representing 
a subordinated goal is shown tower than that repre- 
senting the principal goal. 
232 
recognized by X as final: an SA of type (4) follows 
and the goal is closed, otherwise the reaction is 
not final: an SA of type (5) follows. 
(4) A completing SA reflects the third phase of 
an SR-triad and contains the acceptance of the 
previous positive or negative reaction of the part- 
ner (fig.ld). 
(5) A re-initiating SA ls a re-initiation of 
the goal which means rejection of the reaction to 
the given goal and ptays a double role in the 
dialogue T-structure: such an SA simultaneously 
reflects the third phase of one triad and the first 
phase of the other triad which follows it 
immediately (fig.li). 
In any dialogue complex SAs representing the 
combinations of the enumerated structural types are 
possible. Let us llst some of them: 
(6) An initiating complex SA is a primary ini- 
tiation of the goal with a complication, e.g. with a 
motivation (fig. If). 
(7) A reactive-initiative complex SA is a clo- 
sure of the goal with a complication, for example: 
a) a positive closure with a counter-condition; b) a 
negative closure with a motivation(fig.lg). 
+/- +/- 
Y--1 -1 \] -% 
a b c d i f g 
Fig.1 
If the interaction phases corresponding to the 
types (4) and (5) are expressed implicitly and 
follow by default, then they are designated in the 
structural schemes by a dotted arrow. 
2.3. The elementary SR-triad describes the 
structure of a minimal interaction (e.g. question - 
answer, inducement - refusal etc.). The structural 
schemes given in fig.2 illustrate the positive (a) 
and negative (b) closures of the goal in a 
minimal interaction: 
SAI.X SAI.V SA2.X\[~'\] SAI.X SAI.V SA2.X\[~\] 
(a) (b) 
Fig.2 
SA1.X is an initiation of a certain goal g.X by X. 
SA1.Y is a positive (a) and negative (b) reaction 
of Y to g.X. 
SA2.X is an acceptance of Y's reaction by X (may 
be expressed implicitly). 
In the elementary SR-triad each SA is simple, 
without a complication. SR-triads describe both the 
dialogue fragments and simplest dialogues.The latter 
do not seem to be rare or artificial \[cf. Hundsnur- 
scher,1981J, they are entirely typical, e.g. in 
communicative situation with random unknown part- 
ners: 
(a) SA1.X: What is the time, please? 
SAi.Y: Twenty minutes past four. 
SA2.X: Thanks. 
(b) SA1.X: Have you a cigarette? 
SAI.Y: Unfortunately not. 
SA2.X: Excuse mel 
3. THE TYPES OF RELATIONS BETWEEN SR-TRIADS 
In the general case the dialogue T-structure is 
described by a system of SR-triads which are 
connected with each other by the relations of inter- 
section, imbedding and succession. 
3.1. Let us consider the above types of the 
relations on examples of dialogues with a compli- 
cated T-structure which consists of two SR-triads. 
The intersection of SR-triads represents the T- 
structure of a dialogue containing a motivated 
refusal which is a negative reaction to the goal g.X 
initiated by SA1.X (e.g. to a request). This reac- 
tion includes a motive of the refusal satisfying × 
(see a structural scheme given in fig.3). 
SA1.X SAI.Y SA2.X SA2.Y\[~\] 
Fig.3 
SAt.Y is a negative reaction to g.X with the 
initiation of a reason aimed at the component C of 
the SAI.X content. 
SA2.X is a positive reaction to the reason with 
the removal of the goal g.X. 
SA2.Y is an acceptance of X's reaction to the 
reason by Y (possibly by default). 
In any real dialogue the refusal may be 
expressed indirectly, i.e. SAI.Y contains only the 
reason for non-performing the action to which Y is 
insured. 
The imbeddin~ of SR-triads characterizes the T- 
structure of the dialogue containing an initiating 
complex SA, consists of the initiation of g.X with a 
motivation (e.g. giving an argument for the goal 
g.X). In this case four kinds of the goal closure 
are possible (flg.4(a-d)). 
SA1 .X SA1 .Y SA2.X SA1 .X SA1 .Y SA2.X 
a) b) 
SA1.X SAI.¥ SA2.X SA1.X SAI.Y SA2.X 
c) d) 
Fig.4 
To illustrate these schemes, let us give the 
following variants of the dialogue:*) 
SA1.X:Would you go with me, it is very necessary 
for me. 
SA\].Y:(a)Agreed,I'm all for it. 
(b}Certainly not and you needn't it. 
(c)I'l go, though I think it unnecessary. 
(d)I know it's necessary for you but I 
shan't go. 
SA2.X:I see. 
The succession of SR-triads describes the 
following cases: 
(a) the dialogue in which goals are initiated 
and closed in turns; the connection between these 
goals is not considered in the T-structure (see 
fig.5). 
~L 
SA1.X SAI.V SA2.X SA2.Y SA3.X 
Fig.5 
SA1.X: What is the time of Krasnodar flight? 
*) Certain artificiality of the examples is due to 
the trend to adduce "pure" Illustrations which are 
not complicated by components beyond the scope of 
the schemes considered. 
233 
SAI.Y: It is at six o'clock in the afternoon. 
SA2.X: How much does the grown-up ticket cost? 
SA2.¥: Sixty five roubles. 
SA3.X: Thanks. 
(b) the dialogue containing the re-initlating 
SA2.X. The re-lnitlating g.X means the rejection of 
V's reaction to g.X, initiated in SA1.X (see fig.6). 
SAI.X SAI.Y SA2.X SA2.Y SAS.X 
Fig.6 
SA1.X: Come here, please. 
SA1.¥: No, I shan't. 
SA2.X: Now then, be quickl 
SA2.Y: Well, I come. 
SA3.X: Good for you! 
3.2. More complicated dialogue T-structures 
represent combinations of the above types of the 
relations between SR-triads. These dialogues include 
intermediate speech acts: each SA initiating a 
subordinated goal opens an intermediate SR-triad 
which complicates the dialogue structure. Thus, for 
example, in a dialogue the speech acts of X, which 
are connected with the principal goal initiated by 
the SAI.X, may be defined as "insisting" or 
"persuading". This is the case if Y "offers resis- 
tance" to X, i.e. Y objects, expresses doubts or 
advances counter-arguments. Let us mention as an 
instance a similar dialogue {see the scheme in 
fig.7). 
SAI.X: Let's go to the lecture about Sldorov. 
SA1.Y: No, I doubt whether it will be 
interesting. 
SA2.X: Well, to be sure, Ivanov will deliver the 
lecture himself, there will be slides and tape- 
records. 
SA2.Y: When does it begin? 
SA3.X: In an hour. 
SA3.V: Agreed. 
SA4.X: All right! 
g~X_ - ...... g.X + .... 
+ 
I ÷ I 
ab abe abc ab ab 
SAI.X SA1.Y SA2.X SA2.V SA3.X SA3.Y SA4.X 
Fig.7 
SAI.X is an initiating g.X (offer}. 
SAI.V is:(a) an attempt to close g.X; 
(b) an initiation of the reason (doubt}. 
SA2.X is:(a) a disagreement with the doubt of Y; 
(b) advancing the counter-argument to 
SA1 .V(b} ; 
(c) re-initiating g.X which means rejec- 
tion of the refusal ,\[by default\]. 
SA2.Y is:\[a) an agreement with the counter-argu- 
ment of X \[by default\]; 
(b) an acceptance of X's disagreement 
\[by default\] ; 
(c) a requirement of supplementary 
information. 
SAS.X is:(a) an acceptance of Y's agreement with 
the counter-argument \[by default\]; 
(b) an answer to Y's requirement. 
SA3.¥ is:(a) an acceptance of X's answer \[by 
default\] ; 
(b} a positive reaction to the principal 
goal g.X (agreement}. 
SA4.X is a completing SA. 
In the above communicative situation the patti- 
234 
cipant X may choose another strategy of attaining 
his principal goal g.X: X may initiate g.g not at 
once but "to pave the way" for it, for instance, to 
let Y know of the lecture, to elucidate the degree 
of his interest and, if necessary, to raise it, i.e. 
to construct a sequence of auxiliary SAs and then to 
advance the principal SA.X(offer). The T-structure 
of a similar dialogue interaction may be represented 
by the scheme in fig.8. 
SAI.K SA1.¥ SAZ.X SA2.Y SAS.X SA3.Y SA4.X 
Fig.8 
4.CONCLUSION 
The above SR-triad, being tim basic structural 
element of language interaction, reflects tile inter- 
connection of SAs according to the type of tile 
"stimulus-reaction" relation. It is of special 
importance for us to distinguish the third phase 
which means that the positive or negative reaction 
of a participant to a certain goal initiated by his 
partner becomes final only after the acceptance 
(approval} of this reaction by the initiator of the 
goal. It should be noted that even a positive clo- 
sure of g.X by Y does not always satisfy X, e.g. a 
prompt agreement is interpreted by X as thoughtless 
and insufficiently reasonable. 
The acceptance of a reaction by a dialogue 
participant may happen at different levels of the 
interaction. Thus if Y gives advice to X's request 
£or an advice (there are no violations at the 
communicative level) but tile content of tile advice 
(so-called extracommunicative information} does not 
satisfy X, then he looks in the content of the 
advice for the reason of initiating a further dis-- 
cussion. Tlle above offered apparatus for describing 
the dialogue T-structure with the use of additional 
means reflecting the character of a connection 
between the goals can, apparently, be applied for 
integral multilevel structural representation of LI. 
In the paper we have limited ourselves to the 
consideration of a dialogue (or a fragment of a 
dialogue} subordinated to one goal, its development, 
achievement or removal. However, except the dialogue 
connectedness with respect to goal, other types of 
SAs connectedness in a dialogue may be discussed, 
e.g. a thematic one. The dynamics of the develop- 
ment of the theme does not always correspond to the 
development of the goal. The introduction of the 
notion of the theme and the investigation of its 
relations to the goals will allow us to represent a 
thematic structure to accompany the T-structure of a 
dialogue or to be combined with it. 
REFERENCES 

Frank D. \[1_981~. Seven sins of Pragmatics: Theses 
about Speech Act Theory, Conversational Analysis, 
Linguistics a. Rhetoric, In: Possibilities and Limi- 
tations of Pragmatics. Amsterdam, 1981, p.225-236. 

Hundsnurscher F. ~ On Insisting. In:Possi- 
bilites and Limitations of Pragmatics. Amsterdam, 
1981, p. 343-358. 

Levinson S.C. ~ The Essential Inade- 
quacies of Speech Act Models of Dialogue. In: Possi- 
bilities and Limitations of Pragmaties. Amsterdam, 
1981, 0.473-492. 

Wunderlich D. ~19800)_:. Methodological Remarks on 
Speech Act Theory. In: Speech Act Theory and Pragma- 
tics. Dordrecht etc.,1980, p.291-312. 
