Computational Linguistics & Phonetics Computational Linguistics & Phonetics Fachrichtung 4.7 Universität des Saarlandes

Distinguished Speakers in Language Science

Thursday, 8 January 2015, 16:15
Conference Room, Building C7.4

Speaker adaptations in spontaneous speech in challenging communicative conditions

Valerie Hazan
Department of Speech, Hearing and Phonetics Sciences
UCL

Communication between speakers is often difficult due to the presence of ‘acoustic barriers’ such as noise or signal degradation, or ‘linguistic barriers’, such as a mismatch in native language. In our work, we have investigated how both adults and children adapt the acoustic–phonetic characteristics of their speech to counter the effects of such communication barriers, and what strategies they use in the case of communication breakdowns. We use the ‘diapix’ spot–the–difference picture task to elicit spontaneous speech dialogues between two gender–matched friends; in order to naturally–elicit clear speech strategies on the part of one of the interlocutors, we place an acoustic barrier on the other speaker in certain test conditions via the use of a three–channel vocoder or by adding multispeaker babble to one of the audio channels.

I will review the findings of our first study carried out with 40 adults and of our second study involving 96 young speakers aged between 9 and 14 years inclusive. In our adult study, we did find evidence of adaptations in all speakers, although there were individual differences in the extent of adaptations made across speakers. We also found evidence that the adaptation strategies made were, to an extent, tailored to the specific type of challenging environment that the speaker was aiming to overcome. Children were also making clear speech adaptations in the challenging conditions although some adaptations varied from adults suggesting a less ‘tailored’ approach. There was also evidence of ongoing differences between adults and children in the conversational speech produced in easy communicative conditions. Overall, we find evidence of some development in speech production skills until mid–adolescence at least.

If you would like to meet with the speaker, please contact Jürgen Trouvain.