DiSCoH (Spoken Dialogue System for Conference
Help) is a goal-oriented, mixed-initiative, human-machine spoken
dialogue system for conference information, developed by researchers
at AT&T Labs, ICSI, and Edinburgh University. The main goal
of of DiSCoH is to collect naturally spoken human-machine dialogs.
All of these dialogues will be logged by DiSCoH using rich representations
of context and all data collected with the system will be released
to the research community. It is first being deployed for the
IEEE/ACL SLT Workshop 2006.
Soon you will be able to call
a toll free number for US callers to access DiSCoH for information
about the conference including locations, dates, paper submission
deadlines, program, venue, paper status, accommodation options
and costs, etc.
The central motivation for this
system is that the lack of large, richly annotated dialogue corpora
from real spoken dialogue applications is a major barrier to progress
in the research community. We believe that such a corpus, for
example annotated with speech acts, user utterance transcriptions,
user intentions, overall task success, etc., will be an essential
resource for research in dialogue management, spoken language
understanding, automatic speech recognition, and related tasks.
We invite you to call the system
and help us to collect the DiSCoH spoken dialogue corpus!
|
