Dialogue as Inter-action

Gemma Bel-Enguix and M. Dolores Jim

enez-L

opez

GRLMC-Research Group on Mathematical Linguistics

Rovira i Virgili University

Pl. Imperial Tarraco, 1

43005 Tarragona, Spain

Abstract. In this paper we introduce a formal model of dialogue based on gram-

mar systems theory: Conversational Grammar Systems (CGS). The model takes

into account ideas from the study of human-human dialogue in order to deﬁne a

ﬂexible mechanism for coherent dialogues that may help in the design of effective

and user-friendly computer dialogue systems. The main feature of the model is to

present an action view of dialogue. CGS model dialogue as an inter-action, this is

a sequence of acts performed by two or more agents in a common environment.

We claim that CGS are able to model dialogue with a high degree of ﬂexibility,

what means that they are able to accept new concepts and modify rules, protocols

and settings during the computation.

1 Introduction

Human-computer interaction (HCI) did not exist as a ﬁeld of scientiﬁc inquiry in the

earliest days of computers because very few people interacted with computers, and

those who did generally were technical specialists. Papers on the topic began to appear

only in the 1960s. As more and more people found themselves using computers for a

broadening variety of tasks, the topic became an important focus of research. HCI has

now been a major area of research in computer science, human factors, engineering

psychology and closely related disciplines.

According to [12], a goal of human factors research with computer systems is to

develop human-computer communication modes that are both error tolerant and easily

learned. Since people already have extensive communication skills through their own

native or natural language, many believe that natural language interfaces can provide the

most useful and efﬁcient way for people to interact with computers. Taking into account

this idea, what we propose in this paper is to start by the study and analysis of spoken

human-human dialogues in order to abstract their main principles and mechanisms and

to apply them to the deﬁnition of a formal model of dialogue that may be used for

designing effective, efﬁcient and user-friendly computer dialogue systems.

Research on dialogue has been largely absent from academic disciplines till the sec-

ond half of this 20th century. Importance of dialogue was discovered by an empirical

discipline, known as Conversation Analysis, that emerged in the early 1960s within the

ﬁeld of ethnomethodology. The main purpose of that research stream –always related

to the names of Sacks, Schegloff and Jefferson– can be stated quite simple: to describe

Bel-Enguix G. and Dolores Jiménez-López M. (2007).

Dialogue as Inter-action.

In Proceedings of the 4th International Workshop on Natural Language Processing and Cognitive Science, pages 181-190

DOI: 10.5220/0002418001810190

 SciTePress

‘technology of dialogue.’ The most important objective of conversation analysis is to ex-

plain procedures used by participants in a dialogue to produce utterances and to make

sense of other people’s talk. Being concerned with talk as a collaborative matter and

with how parties can jointly produce an organised sequence of talk, conversation analy-

sis tries to specify how the consecutive actions that dialogue consists of are related one

to another and how they build up a conversational sequence. Methodology and results

obtained by researchers in the ﬁeld of conversation analysis have revealed as quite use-

ful in the area of human-computer interaction. Computer scientists such as Norman and

Thomas or Robinson have pointed out that utility:

‘Conversation Analysis seems to us to offer the possibility of the provision

of comprehensive and secure design information based on a coherent view of

interaction, although representing an investigative paradigm quite different to

those currently employed in Human-Computer Interaction research.’ [11].

‘The ﬁndings of one particular form of ethnomethodological work, that of

Conversational Analysis, seem prima facie to be directly relevant to human-

computer interaction.’ [14].

The model we introduce in this paper is based on the theory of grammar systems

and takes into account ideas from the study of human-human dialogue in order to de-

ﬁne a ﬂexible mechanism for coherent dialogues. The aim of the model is to see how

productive can be to reproduce in human-computer dialogues details of natural conver-

sations between people. In order to fulﬁl that goal, we use in our formal model ideas,

techniques and procedures that have been proposed to account for human dialogue. The

main feature of this model is to present an action view of dialogue. Therefore, next sec-

tion will be devoted to overview some action-based approaches to language. Section 3

will introduce Conversational Grammar Systems as a formal-language-model that de-

ﬁnes dialogue as Inter-Action. Last section present some ﬁnal remarks and directions

for future work.

Throughout the paper, we assume that the reader is familiar with the basics of formal

language theory, for more information see [15].

2 Dialogue as Inter-action

Within a philosophical tradition begun by Austin [3], dialogue is viewed as a sequence

of speech acts, uttered by each party to achieve certain goals. He observes that there

exists a type of utterances that do not describe or report anything at all, but that their

uttering is the doing of an action. He calls this special type of utterances performative

sentences, in order to stress the idea that the issuing of the utterance is the performing

of an action and not just the saying of something (as is the case of constative sentences).

After having postulated the existence of performative sentences that cannot be said to

be ‘true’ or ‘false,’ but that can be qualiﬁed at most as ‘happy’ or ‘unhappy,’ Austin

observes that in any utterance we can individuate three different types of acts, being

one of them the act of doing something while uttering the sentence.

182

By introducing the idea of illocutionary act, Austin opens what has been a very

inﬂuential theory, namely theory of speech acts. However, that theory would not have

had the repercussion it has actually had without the ﬁgure of John Searle. Work done

by this author in the ﬁeld of speech acts is considered as the systematic development

and continuation of Austin’s

Speech act theory considers the interactive use of language to be of primary impor-

tance. According to [6], speech act theory has been a major source of inspiration for

all action-based approaches to language, and has been fruitful both in the development

of pragmatics and as a conceptual framework for thinking about human computer di-

alogue. Action view of dialogue is perfectly resumed in Searle’s sentence: ‘Talking is

performing acts according to rules.’ [16].

An important action theory of language is Dynamic Interpretation Theory (DIT)

introduced by Bunt [5]. In DIT, dialogues are viewed in an action perspective. Language

is considered a tool to perform context-changing actions. According to DIT, a dialogue

can be analysed in terms of combinations of actions called dialogue acts deﬁned as:

‘Functional units used by the speaker to change the context.’ [5].

Many authors have defended the idea that to use language is to perform acts ac-

cording to rules. Next to speech act theory or Bunt’s DIT, we can ﬁnd thesis as the one

presented by Clark [7] who views language use –and, therefore, dialogue– as a joint

action, deﬁning joint action as ‘One that is carried out by an ensemble of people act-

ing in coordination with each other.’ For this researcher, ‘What people do in arenas of

language use is to take actions.’.

In a similar fashion, in [9] it is claimed that analysis of dialogue ought to be based

on a theory of collective action. These authors take language as action and study those

aspects of language use which can be explained following general principles of coop-

erative interaction. Also for Sharrock & Anderson, the primary characteristic of the

utterances conversation analysis deals with is ‘often less that they are verbal actions,

but that they are actions.’ [17].

Another general action-based approach to language has been developed by All-

wood and co-workers and has been called Communicative Activity Analysis [1] [2]. Like

speech act theory, Allwood’s approach takes the view that communication is action and

provides a conceptual analysis of action, social activity and ethics in communication

with considerable depth and generality.

The above are just few examples of the action view of language use. All of them

share the idea of deﬁning a dialogue as to perform actions in a speciﬁc context. In order

to apply these ideas to the design of dialogue systems, we need a formalized theory

that takes into account both general principles of natural language dialogue and, of

course, this generalized view of dialogue as action. In the next section, we provide a

formal model that tries to capture these ideas by using a formal-language-theoretical

framework.

3 Conversational Grammar Systems: An Inter-action Model

The model we introduce here is based on Grammar Systems Theory. Grammar systems

can be characterized as a device where agents perform actions according to rules.

183

Grammar systems theory is a consolidated and active branch in the ﬁeld of formal

languages [8] that provides syntactic models for describing multi-agent systems at the

symbolic level, using tools from formal grammars and languages. Grammar systems

theory has been widely investigated and nowadays constitutes a well-developed formal

theory that presents several advantages with respect to classical models. However, being

a branch of formal languages, researchers in the ﬁeld of grammar systems have concen-

trated mainly on theoretical aspects. Roughly speaking, a grammar system is a set of

grammars working together, according to a speciﬁed protocol, to generate a language.

Notice that while in classical formal language theory one grammar (or automata) works

individually to generate (or recognize) one language; here, instead, we have several

grammars working together in order to produce one language.

While grammar systems are related to Artiﬁcial Intelligence, a subﬁeld of the theory,

–the so-called eco-grammar systems– is closely related to Artiﬁcial Life. Eco-grammar

systems provide a syntactical framework for eco-systems, this is, for communities of

evolving agents and their interrelated environment. Brieﬂy, an eco-grammar system is

deﬁned as a multi-agent system where different components, apart from interacting

among themselves, interact with a special component called ‘environment’ [13].

Here we introduce a new model: Conversational Grammar Systems (CGS). CGS

are multi-agent systems based on grammar systems, speciﬁcally in the so-called eco-

grammar systems. Conversational grammar system offer a framework with a high de-

gree of ﬂexibility, what means that they are able to accept new concepts and mod-

ify rules, protocols and settings during the computation. Evolution and action are in-

volved in a consistent way in environment/contexts, while inter-action of agents with

the medium is constant.

According to the idea that dialogue ‘can be understood as the sustained production

of chains of mutually-dependent acts, constructed by two or more agents each moni-

toring and building on the actions of the other.’ [10], conversational grammar systems

intend to describe dialogue as a sequence of context-change-actions allowed by cur-

rent environment and performed by two or more agents. Therefore, in conversational

grammar systems we understand dialogue as inter-action, this is a sequence of acts

performed by two or more agents in a common environment.

In what follows we introduce the formal deﬁnition of our model.

Deﬁnition 1 A Conversational Grammar System (CGS) of degree n, n ≥ 2, is an (n +

1)-tuple:

Σ = (E, A

, ..., A

where:

– E = (V

, P

• V

: an alphabet;

• P

: a ﬁnite set of rewriting rules on V

– A

= (V

, P

, R

, ϕ

, ψ

, π

, ρ

) , 1 ≤ i ≤ n,

• V

: an alphabet;

• P

: a ﬁnite set of rewriting rules on V

;

184

• R

: a ﬁnite set of rewriting rules on V

;

• ϕ

: V

∗

→ 2

;

• ψ

: V

∗

× V

→ 2

;

• π

: the start condition;

• ρ

: the stop condition;

• π

and ρ

: predicates on V

∗

. We can deﬁne the following special types of pred-

icates. We say that predicate σ on V

∗

is of:

∗ Type (a) iff σ(w) = true for all w ∈ V

∗

;

∗ Type (rc) iff there are two subsets R and Q of V

and σ(w) = true iff w

contains all letters of R and w contains no letter of Q;

∗ Type (K) iff there are two words x and x

′

over V

and σ(w) = true iff x

is a subword of w and x

′

is not a subword of w;

∗ Type (K

′

) iff there are two ﬁnite subsets R and Q of V

∗

and σ(w) = true

iff all words of R are subwords of w and no word of Q is a subword of w;

∗ Type (C) iff there is a regular set R over V

and σ(w) = true iff w ∈ R.

The items of the above deﬁnition have been interpreted as follows: a) E represents

the environment described at any moment of time by a string w

, over alphabet V

called the state of the environment. The state of the environment is changed both by

its own evolution rules P

and by the actions of the agents of the system, A

, 1 ≤

i ≤ n. b) A

, 1 ≤ i ≤ n, represents an agent. It is identiﬁed at any moment by a

string of symbols w

, over alphabet V

, which represents its current state. This state

can be changed by applying evolution rules from P

, which are selected according to

mapping ϕ

and depend on the state of the environment. A

can modify the state of

the environment by applying some of its action rules from R

, which are selected by

mapping ψ

and depend both on the state of the environment and on the state of the

agent itself. Start/Stop conditions of A

are determined by π

and ρ

, respectively. A

starts/stops its actions if context matches π

and ρ

. Start/stop conditions of A

can be

of different types: (a) states that an agent can start/stop at any moment. (rc) means that

it can start/stop only if some letters are present/absent in the current sentential form.

And (K), (K

′

) and (C) denote such cases where global context conditions have to be

satisﬁed by the current sentential form.

CGSs intend to describe dialogue as a sequence of context-change-actions allowed

by the current environment and performed by two or more agents. In this view, an action

is deﬁned as the application of a rule on the environmental string:

Deﬁnition 2 By an action of an active agent A

in state σ = (w

; w

, w

, . . . , w

) we

mean a direct derivation step performed on the environmental state w

by the current

action rule set ψ

, w

) of A

Deﬁnition 3 A state of a CGS Σ = (E, A

, . . . , A

), n ≥ 2, is an n + 1-tuple:

σ = (w

; w

, . . . , w

where w

∈ V

∗

is the state of the environment, and w

∈ V

∗

, 1 ≤ i ≤ n, is the state

of agent A

185

AGENTS

...

evolution

rules

...

description

...

action

rules

description

ENVIRONMENT

evolution

rules

- - -

?? ?? ??

? ? ?

6666 66

Fig.1. Conversational Grammar Systems.

This rule is applied by an active agent and it is a rule selected by ψ

, w

Deﬁnition 4 An agent A

is said to be active in state σ = (w

; w

, w

, . . . , w

) if the

set of its current action rules, that is, ψ

, w

), is a nonempty set.

Since dialogue in CGS is understood in terms of context changes , we have to deﬁne

how the environment passes from one state to another as a result of agents’ actions:

Deﬁnition 5 Let σ = (w

; w

, . . . , w

) and σ

′

= (w

′

; w

′

, . . . , w

′

) be two states of

a CGS Σ = (E, A

, . . . , A

). We say that σ

′

arises from σ by a simultaneous action

of active agents A

, . . . , A

, where {i

, . . . , i

} ⊆ {1, . . . , n}, i

6= i

, for j 6= k,

1 ≤ j, k ≤ r, onto the state of the environment w

, denoted by σ

=⇒

′

, iff:

– w

= x

. . . x

and w

′

= y

. . . y

, where x

directly derives y

by using

current rule set ψ

, w

) of agent A

, 1 ≤ j ≤ r;

– there is a derivation:

= w

=⇒

∗

=⇒

∗

=⇒

∗

. . .

=⇒

∗

= w

′

such that, for 1 ≤ j ≤ r, π

j−1

) = true and ρ

) = true. And for f ∈ {t, ≤

k, ≥ k} the derivation is:

= w

=⇒

. . .

=⇒

= w

′

such that, for 1 ≤ j ≤ r, π

j−1

) = true

, and

– w

′

= w

, 1 ≤ i ≤ n.

In this latter case the stop condition ρ

) = true is replaced by the stop condition given the

f-mode.

186

However, in the course of a dialogue, agents’ states are also modiﬁed and the envi-

ronmental string is subject to changes due to reasons different from agents’ actions. So,

in order to complete our formalization of dialogue development, we add the following

deﬁnition:

Deﬁnition 6 Let σ = (w

; w

, . . . , w

) and σ

′

= (w

′

; w

′

, . . . , w

′

) be two states of a

CGS Σ = (E, A

, . . . , A

). We say that σ

′

arises from σ by an evolution step, denoted

by σ

=⇒

′

, iff the following conditions hold:

– w

′

can be directly derived from w

by applying rewriting rule set P

;

– w

′

can be directly derived from w

by applying rewriting rule set ϕ

), 1 ≤ i ≤

In CGS, the development of dialogue implies that both the state of the environment

and state of agents change. Such changes take place thanks to two different types of pro-

cesses: action steps and evolution steps. By means of the former, active agents perform

actions on the environmental string modifying its state; the latter imply the reaction of

context and agents which, according to the changes produced by agents’ actions, mod-

ify their states. So, action steps and evolution steps alternate in the course of dialogue.

At the end, what we have is a sequence of states reachable from the initial state by

performing, alternatively, action and evolution derivation steps:

Deﬁnition 7 Let Σ = (E, A

, . . . , A

) be a CGS and let σ

be a state of Σ. By a state

sequence (a derivation) starting from an initial state σ

of Σ we mean a sequence of

states {σ

}

∞

i=0

, where:

– σ

=⇒

i+1

, for i = 2j, j ≥ 0; and

– σ

=⇒

i+1

, for i = 2j + 1, j ≥ 0.

Deﬁnition 8 For a given CGS Σ and an initial state σ

of Σ, we denote the set of state

sequences of Σ starting from σ

by Seq(Σ, σ

The set of environmental state sequences is:

Seq

(Σ, σ

) = {{w

}

∞

i=1

| {σ

}

∞

i=0

∈ Seq(Σ, σ

), σ

= (w

; w

, . . . , w

)}.

The set of state sequences of the j-th agent is deﬁned by:

Seq

(Σ, σ

) = {{w

}

∞

i=1

| {σ

}

∞

i=0

∈ Seq(Σ, σ

), σ

= (w

; w

, . . . , w

)}.

Now, we associate certain languages with an initial conﬁguration:

Deﬁnition 9 For a given CGS Σ and an initial state σ

of Σ, the language of the

environment is:

(Σ, σ

) = {w

∈ V

∗

| {σ

}

∞

i=0

∈ Seq(Σ, σ

), σ

= (w

; w

, . . . , w

)}.

and the language of j-th agent is:

(Σ, σ

) = {w

∈ V

∗

| {σ

}

∞

i=0

∈ Seq(Σ, σ

), σ

= (w

; w

, . . . , w

)}.

for j = 1, 2, . . . , n.

Two important selection techniques in dialogue are the turn-taking system and the

adjacency pairs. If we want to provide a formal language account of turn-taking, we

should focus on the most important traits of this phenomenon, and make it susceptible

to formalization. In order to do so, we deﬁne different derivation modes that control

how long an agent can act in the environmental state:

187

Deﬁnition 10 Let Σ = (E, A

, ..., A

) be a CGS. And let w

= x

...x

and w

′

...y

be two states of the environment. Let us consider that w

′

directly derives from

by action of active agent A

, 1 ≤ i ≤ n, as shown in Deﬁnition 5. We write that:

=⇒

≤k

′

iff w

=⇒

≤k

′

, for some k

′

≤ k;

=⇒

≥k

′

iff w

=⇒

≤k

′

, for some k

′

≥ k;

=⇒

∗

′

iff w

=⇒

′

, for some k;

=⇒

′

iff w

=⇒

∗

′

and there is no z 6= y with y

=⇒

∗

In words, ≤ k-derivation mode represents a time limitation where A

can perform

at most k successive actions on the environmental string. ≥ k-derivation mode refers

to the situation in which A

has to perform at least k actions whenever it participates

in the derivation process. With ∗-mode, we refer to such situations in which agent A

performs as many actions as it wants to. And ﬁnally, t-derivation mode represents such

cases in which A

has to act on the environmental string as long as it can.

One way of getting transitions with no gap and no overlap in CGS is to endow

agents with an internal control that contains start/stop conditions that allow agents to

recognize places where they can start their activity, as well as places where they should

stop their actions and give others the chance to act. This is, start/stop conditions help

agents to recognize transition relevance places, i.e. places where speaker change oc-

curs. Start/stop conditions have been formally deﬁned in Deﬁnition 1.

It seems quite common in talk exchanges to ﬁnd paired actions. Notions such as

adjacency pairs, reactive pressures, discourse expectations etc. intend to account for

the fact that utterances produced in dialogue are somehow determined and constrained

by preceding utterances in the talk exchange. Mapping ψ

, w

) fulﬁls in CGS a

function analogous to the one carried out by all the above notions in their respective

conversational models. This mapping establishes which actions are allowed for agent

at any given moment.

Closing a dialogue implies that participants stop their conversational activity be-

cause they have reached their goal in the talk exchange. For deciding when the compu-

tation terminates, we have to determine which string is to be considered as the reference

point to signal the end of the derivation. We can identify at least three different styles

of closing derivation process in CGS:

Deﬁnition 11 Let Σ = (E, A

, ..., A

) be a CGS as in Deﬁnition 1. Derivation in Σ

terminates in:

– Style (ex) iff for A

, ..., A

, ∃A

: w

∈ T

, 1 ≤ i ≤ n;

– Style (all) iff for A

, ..., A

, ∀A

: w

∈ T

, 1 ≤ i ≤ n;

– Style (one) iff for A

, ..., A

, A

: w

∈ T

, 1 ≤ i ≤ n.

According to the above deﬁnition, a derivation process ends in style (ex) if there is

some agent A

that has reached a terminal string. It ends in style (all) if every agent in

the system has a terminal string as state. And it ﬁnishes in style (one) if there is one

distinguished agent whose state contains a terminal string. Styles (all), (ex) and (one)

might account for three different ways of closing a dialogue.

The following simple example illustrates how CGS work.

188

Example 1 Consider the following CGS: Σ = (E, A

, A

), where:

– E = (V

, P

• V

= {a, x, y};

• P

= {a → b

, b → a

, x → x, y → y}.

– A

= (V

, P

, R

, ϕ

, ψ

, π

, ρ

) with:

• V

= {c};

• P

= {c → c}; R

= {a → x};

• ϕ

(w) = P

for every w ∈ V

∗

;

• ψ

(w; u) = R

for w ∈ {a, x, y}

∗

and u = c, otherwise ψ

(w; u) = ∅;

• π

= true for all w ∈ V

∗

; ρ

= true for all w ∈ V

∗

– A

= (V

, P

, R

, ϕ

, ψ

, π

, ρ

) with:

• V

= {d};

• P

= {d → d}; R

= {b → y};

• ϕ

(w) = P

for every w ∈ V

∗

;

• ψ

(w; v) = R

for w ∈ {b, x, y}

∗

and v = d, otherwise ψ

(w; v) = ∅;

• π

= true for all w ∈ V

∗

; ρ

= true for all w ∈ V

∗

, P

and P

contain rules of an 0L system applied in a parallel way. Rules in R

and R

are pure context-free productions applied sequentially. Let us suppose that the

system is working in the arbitrary mode ∗. And let us take σ

= (a

; c, d) as the initial

state of Σ. Then, a possible derivation in Σ is the following one:

; c, d)

=⇒

∗

x; c, d)

=⇒

∗

x; c, d)

=⇒

∗

(yb

x; c, d)

=⇒

∗

(ya

x; c, d)

=⇒

∗

(ya

x; c, d)

=⇒

∗

. . .

Notice, that we alternate action and evolution steps. At every action step one of the

agents rewrites one symbol of the environmental state, while in evolution steps both

environmental and agents’ states are rewritten according to 0L rules.

4 Final Remarks and Future Work

In this paper we have introduced a formal model of dialogue based on grammar sys-

tems. Conversational grammar systems are able to model dialogue with a high degree

of ﬂexibility, what means that they are able to accept new concepts and modify rules,

protocols and settings during the computation. Evolution and action are involved in a

consistent way in environment/contexts, while interaction of agents with the medium is

constant. CGS present some advantages to account for dialogue: a) generation process

is highly modularised by a distributed system of contributing agents; b) it is contextual-

ized, linguistic agents re-deﬁne their capabilities according to context conditions given

by mappings; c) and emergent, it emerges from current competence of the collection of

active agents.

Moreover, we claim that CGS provides a powerful framework for formalizing any

type of inter-action, both among agents and among agents and the environment. Of

course, a topic where context and interaction among agents is essential is the ﬁeld of

189

dialogue modelling and its applications to the design of effective and user-friendly com-

puter dialogue systems where we think our model can be directly applied.

Finally, it seems this system is quite easy to implement, due to the simplicity of

the formalism and the computational background of the multi-agent theory we use.

Achieving a valid and simple computational implementation of this formal framework

is the major research line for the future. A simple example of implementation of a

variant of this model can be found in [4].

References

1. Allwood, J.: Linguistic Communication as Action and Cooperation. Gothenburg Mono-

graphs in Linguistics 2, G

oteborg University (1976)

2. Allwood, J.: Obligations and Options in Dialogue. Think 3 (1994) 9–18.

3. Austin, J.L.: How to Do Things with Words. Harvard University Press, Cambridge (1962).

4. Bel-Enguix, G., Grando, A., Jim

enez-L

opez, M.D.: A grammatical framework for modlling

multi-agent dialogues. In: Shi, Z. & Sadananda, R. (eds.): Agent Computing and Multi-

Agent Systems: 9th Paciﬁc Rim International Workshop on Multi-Agents, PRIMA 2006.

LNCS 4088, Springer, Berlin (2006) 10–21

5. Bunt, H.C.: Context and Dialogue Control. Think 3 (1994) 19–30.

6. Bunt, H.: Dialogue Pragmatics and Context Speciﬁcation. In: Bunt, H., Black, W. (eds.):

Abduction, Belief and Content in Dialogue. John Benjamins, Amsterdam (2000) 81–150.

7. Clark, H.H.: Using Language. Cambridge University Press, Cambridge (1996).

8. Csuhaj-Varj

u, E., Dassow, J., Kelemen, J., P

aun, Gh.: Grammar Systems: A Grammatical

Approach to Distribution and Cooperation. Gordon and Breach, London (1994)

9. Korta, K., Larrazabal J.: Dialogue as Action. In: Korta, K., Larrazabal J. (eds.): Semantics

and Pragmatics of Natural Language: Logical and Computational Aspects. Proceedings of

the Donostia-Toulouse 1993 Workshop. ILCLI, San Sebasti

an (1993) 7–23

10. Levinson, S.C.: Pragmatics. Cambridge University Press, Cambridge (1983).

11. Norman, M., Thomas, P.: The Very Idea. Informing HCI Design from Conversation Analysis.

In: Luff, P., Gilbert, N., Frohlich, D. (eds.): Computers and Conversation. Academic Press,

London (1990) 51–65.

12. Ogden, W.C., Bernick, Ph.: Using Natural Language Interfaces. In: Helander, M.G., Lan-

dauer, T.K., Prabhu, P.V. (eds.): Handbook of Human-Computer Interaction. Elsevier, Ams-

terdam (1997) 137–161.

13. P

aun, Gh. (ed.): Artiﬁcial Life: Grammatical Models. Black Sea University Press, Bucharest

(1995).

14. Robinson, H.: Towards a Sociology of Human-Computer Interaction. A Software Engineer’s

Perspective. In: Luff, P., Gilbert, N., Frohlich, D. (eds.): Computers and Conversation. Aca-

demic Press, London (1990) 39–49.

15. Rozenberg, G., Salomaa, A.: Handbook of Formal Languages. Springer, Berlin (1997)

16. Searle, J.R.: Speech Acts. An Essay in the Philosophy of Language. Cambridge University

Press, Cambridge (1969).

17. Sharrock, W., Anderson, B.: Epilogue: The Deﬁnition of Alternatives: Some Sources of Con-

fusion in Interdisciplinary Discussion. In: Button, G., Lee, J.R.E. (eds.): Talk and Social

Organization. Multilingual Matters LTD, Clevedon (1987) 290–321.

190