Graphics Reference
In-Depth Information
CHAPTER 1
Speech Technology and
Conversational Activity in
Human-Machine Interaction
Nick Campbell
1. Introduction
Over the past several years, great strides have been made in both
speech synthesis and speech recognition technology, but so far,
comparatively little work has been carried out on the higher-level
integration of these components to enable them to be used for a
more naturally interactive form of conversation between people and
machines (or between people by way of machines). This chapter
discusses the need for a more proactive form of speech processing,
presents some ideas related to the types of speech structure that
might need to be modeled for facilitating more spontaneous
conversational interactions, and describes some of our recent
work in this direction at the Speech Communication Lab of the Centre
for Language and Communication Studies at Trinity College Dublin.
Most spoken dialogue interfaces are implicitly asynchronous, as
they require a text-based processing step between each utterance.
This form of processing enforces a 'ping-pong' type of interaction
with explicit turn-taking. The incoming speech is first recognized, or
converted into text, then parsed and processed, and the resulting text
output is then rendered as speech output by a synthesizer.
