5.0444 Workshop: Natural Language Dialogue Systems (1/72)

Elaine Brennan & Allen Renear (EDITORS@BROWNVM.BITNET)
Mon, 11 Nov 1991 18:53:33 EST

Humanist Discussion Group, Vol. 5, No. 0444. Monday, 11 Nov 1991.

Date: Wed, 6 Nov 91 11:52:31 -0500
From: walker@flash.bellcore.com (Don Walker)
Subject: Workshop on Natural Language Dialogue Systems

ANLP-92 Workshop

CALL FOR PARTICIPATION

Trento, Italy, 30 March 1992

EMPIRICAL MODELS AND METHODOLOGY FOR NATURAL LANGUAGE DIALOGUE SYSTEMS

jointly organized by

Lars Ahrenberg Nils Dahlback Arne Jonsson
Department of Computer and Information Science
Linkoping University

There is an increasing recognition of the need to put theories and
systems within the field of Natural Language Processing on a firm
empirical ground (cf. the views of the panel on Language Engineering:
The Real Bottleneck of Natural Language Processing at COLING-88).
Researchers on various aspects of discourse, especially, take pains
to base, or at least illustrate, their theoretical developments
with some kind of corpus data. There is quite often a problem,
however, in determining what data are relevant for a specific domain
or application and in deciding what conclusions the data warrant.

In this workshop we want to focus on data used for the design of
natural language interfaces and other dialogue systems accepting
natural language input. In this area we know of three basic sources
for the corpora used: (i) interactions with computers using existing
NLIs, (ii) interactions with computers using simulated NLIs (Wizard
of Oz-studies), and (iii) human-human dialogues collected and/or
analyzed with the purpose of developing computational theories of
discourse.

Several studies of this kind have been published in recent years.
However, the use of these different kinds of corpus data raises a
number of issues in need of clarification, theoretical as well as
methodological: e.g. what are the advantages and disadvantages of
corpora collected with these different methods; what is the
generalizability of the results obtained; what methodological
pitfallss have been found, what means to avoid them, and so on.
Behind many of these questions lies the issue of the extent to
which natural language as used in human-computer interaction is to
be seen as a distinct sublanguage, and to what extent there are
aspects common to all kinds of dialogue.

The workshop will adress both the methodological and the theoretical
issues raised above, but with the focus on the methodological ones.
Its aim is to give active researchers in the field an opportunity
to exchange experiences of and opinions on different
practical/methodological topics such as those mentioned above, and
others. If possible, demonstrations of software for the collection
and analysis of these kind of data will be demonstrated.

Those interested in making a presentation should submit an abstract
of 500-1000 words. The abstract should include a description of
the participant's current research in the area and point to longer
papers on the subject (if available).

Those interested only in participating should submit a statement of
their research interests together with a list of related publications.

Submissions, by e-mail or surface mail, should arrive no later than
20 January 1991 to the address below.

Nils Dahlback
Department of Computer and Information Science
Linkoping University
S-581 83 Linkoping, SWEDEN
email: nda@ida.liu.se

Notification of acceptance will be mailed by the end of January 1992.