Toward Constructing a Linguistic Model for the Process of Spoken
Dialogue
<P>
Hiroya FUJISAKI, Sumio OHNO
<P>
Department of Applied Electronics, Science University of Tokyo<P>
2641 Noda, 278 Japan<P>
e-mail: fujisaki@te.noda.sut.ac.jp
<P>
Spoken language is clearly the primary means of human communication,
and will no doubt be the major means of man-machine communication in
the future.  Compared with the written language, however, messages of
the spoken language are less well-formed, with frequent occurrence of
omissions, repetitions, word-order changes, errors, etc.  Yet these
so-called ill-formed features do not seem to present any obstacles to
a human-to-human spoken dialogue.  This leads us to believe that these
features possess their own reason of existence and are actually
utilized by the participants of a spoken dialogue to facilitate smooth
and efficient information exchange.  From this point of view, the
present study analyzes human spoken dialogues and then tries to
construct a model for the process of a spoken dialogue, with an aim of
its utilization in human-machine dialogues.  The followings are
summaries of our preliminary analysis.

\medskip
Prerequisites to a Spoken Dialogue.

Spoken dialogue is a purposeful exchange of information through spoken
language among agreed participants.  In many cases, the flow of
information is globally unidirectional, but is often locally
bidirectional.  Participants generally possess a certain amount of
knowledge and expectation on the knowledge, ability, and intentions of
the other participants.

\medskip
Basic Constraints of a Spoken Dialogue.

\noindent
(1) Parsimony --- For the purpose of prompt communication, information
that is judged to be redundant is not transmitted, hence frequent use
of deixes, ellipses, etc.

\noindent
(2) Real-time production --- Prompt, real-time exchange of information
does not allow the speaker to construct the whole message in advance,
hence frequent occurrence of hesitations, filler sounds, re-starts,
etc.  These irregularities, however, do reflect the process of
information processing on the part of the speaker, and may help
correct decoding of the received message on the part of the listener.

\medskip
Principles and Methods for a Smooth Spoken Dialogue.

\noindent
(1) Principle of cooperation --- The speaker selects the information
which is supposed to be most relevant for the listener, and uses
expressions that are supposed to be easily understood by the listener.
The listener also selects, among possible alternatives, the semantic
interpretation that is supposed to be meant by the speaker, and infers
the speaker's intention that may not be explicitly stated.

\noindent
(2) Principle of reliability --- In order to ensure the overall
reliability of communication, the speaker uses specific methods for
the following purposes:<P>
a. Drawing attention using prominence and repetition.<P>
b. Confirmation of the listener's comprehension.<P>
c. Indication of the speaker's own information processing stages.<P>
d. Indication of retention, transfer, and request of speaking turns.<P>
e. Repair of errors both on the part of the speaker and of the listener.
<P>
Keywords: spoken dialogue, linguistic model, ill-formedness, parsimony,
real-time production, principle of cooperation, principle of reliability<P>