Turn Detection
The identification of speaker-turn boundaries or conversational transitions in continuous audio.
Speech Processing Context
Turn detection decides when an interaction unit has ended so a streaming system can act without waiting unnecessarily. VAD state, pause duration, linguistic completion, and speaker cues can be combined; an early decision truncates utterances, while a late decision increases interaction latency.
Modeling Boundary
Turn detection and speaker diarization answer different questions. Turn detection locates interaction boundaries; diarization assigns speech regions to speakers.