Multiutterance

Multiutterance is a term used in linguistics and speech technology to describe a sequence in which a speaker produces multiple utterances in close succession, either within a single turn or as a rapidly linked series across turns. The unit may consist of separate clauses, phrases, or sentences that are syntactically connected or loosely related, and it can include repairs, elaborations, or reiterations.

In discourse analysis, multiutterance is examined to understand how speakers manage turn-taking, coherence, and emphasis. Prosody, pacing, and pausing contribute to signaling the boundary between utterances, and discourse markers can guide listener interpretation across the sequence. The phenomenon often arises in spontaneous conversation, where a speaker quickly adds information, corrects a previous statement, or presents a chain of related ideas.

In corpus annotation and speech technology, identifying multiutterance segments involves detecting utterance boundaries, which may be indicated by punctuation, intonation, or explicit markers. Automated methods use acoustic features, textual cues, and discourse signals to segment and label each utterance within a sequence. Accurate annotation supports tasks such as transcription, discourse analysis, and model training for conversational systems.

Applications of multiutterance analysis include improving the naturalness of conversational agents, enhancing transcription fidelity, and refining speech recognition in interactive systems. Generating natural-looking multiutterance sequences can help bots convey nuance, emphasis, and conditional or elaborative content in dialogue.

See also: turn-taking, discourse markers, prosody, utterance, segmentation.