We present techniques for
the incremental interpretation and prediction of utterance meaning in dialogue systems.
These techniques open possibilities for systems to initiate responsive overlap behaviors
during user speech, such as interrupting, acknowledging, or completing a user's
utterance while it is still in progress. In an implemented system, we show that
relatively high accuracy can be achieved in understanding of spontaneous utterances
before utterances are completed. Further, we present a method for determining when a
system has reached a point of maximal understanding of an ongoing user utterance, and
show that this determination can be made with high precision. Finally, we discuss a
prototype implementation that shows how systems can use these abilities to strategically
initiate system completions of user utterances. More broadly, this framework facilitates
the implementation of a range of overlap behaviors that are common in human dialogue,
but have been largely absent in dialogue systems.