Cross-linguistic study of the production of turn-taking cues in American English and Argentine Spanish

PABLO BRUSCO; GRAVANO, AGUSTÍN; JUAN MANUEL PÉREZ

Estocolmo

Conferencia; Interspeech 2017; 2017

International Speech Communication Association

We present the results of a series of machine learning experi-ments aimed at exploring the differences and similarities in theproduction of turn-taking cues in American English and Argen-tine Spanish. An analysis of prosodic features automaticallyextracted from 21 dyadic conversations (12 En, 9 Sp) revealedthat, when signaling Holds, speakers of both languages tend touse roughly the same combination of cues, characterized by asustained final intonation, a shorter duration of turn-final inter-pausal units, and a distinct voice quality. However, in speechpreceding Smooth Switches or Backchannels, we observe theexistence of the same set of prosodic turn-taking cues in bothlanguages, although the ways in which these cues are combinedtogether to form complex signals differ. Still, we find that thesedifferences do not degrade below chance the performance ofcross-linguistic systems for automatically detecting turn-takingsignals. These results are relevant to the construction of multi-lingual spoken dialogue systems, which need to adapt not onlytheir ASR modules but also the way prosodic turn-taking cuesare synthesized and recognized.