A human Dialogue is a human to human interaction.
turn
each contributino to a conversation is called a “turn”, which contains a sentence, multiple sentences, or a single word
turn-taking
- when to take the floor
- who takes the floor
- what happens during interruptions?
barge-in
barge-in is the property to allow the user to interrupt the system
end-pointing
deciding when a human has stopped talking, compute, etc.
speech-act
each turn is actually an “action” performed by the user
- constatives: committing the speaker to something being the case (answering, denying)
- directives: ask the addressee to do something (advising, ordering)
- com missives: commuting the speaker to future action (planning, voving)
- acknowledgement: reflecting the speaker’s attitude for something (apologizing, greeting, etc.)
common ground
grounding is the problem of acknowledging and reflecting the state of interaction; such as the elevator lighting up when pressed.
acknowledgements and repeats is a way of grounding.
we need to make sure that the system acknowledges user interaction
adjacency pairs
- question => answer
- proposal => acceptance/rejection
- complements => downplay
two-pair composition maybe interrupted or separated by a sub-dialogue
conversational initiative
Sometimes, such as during interviews, only one agent has initiative. This is not true most of the time during human-human interactions.
mixed initiative is hard to achieve, usually we make dialogue systems as passive environments—only user and system understanding.