Abstract

Two studies of spoken telephone number transfers––in which one participant communicates a number to the other––are reported. The data comprised transcripts of telephone conversations between callers and operators; for the second corpus audio recordings were also available. In most cases the caller was giving the number to the operator, and in these dialogues a chunked echo protocol was found to be very common, with the operator repeating each chunk of one or more digits to the caller for confirmation before the next chunk was given. Errors in speaking and in recognition were corrected efficiently within this protocol. The observations support a model of dialogue in which a single utterance unit can perform multiple dialogue acts and in which discourse units can have hierarchical structure. Examination of the audio recordings showed that there was usually very little silence, and sometimes a slight overlap, between conversational turns. Various prosodic phenomena were noted as contributing to the turn-taking and grounding processes. Implications for automated dialogue systems are discussed.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call