Phonetic Convergence in Human-Computer Interaction

• Applicant: Professor Dr. Bernd Möbius, since 1/2019
• Subject Area: General and Comparative Linguistics, Experimental Linguistics, Typology, Non-European Languages ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: from 2015 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 278805297

The proposed research project aims at the analysis, quantification,modeling, and evaluation of phonetic convergence in human-computerinteraction. Phonetic convergence is defined as an increase insegmental and suprasegmental similarities between two speakers,presumably grounded in spontaneous phonetic adoption of speechcharacteristics of the interlocutor. Building on research onconvergence in human-human-communication, this project will develop,implement, and evaluate a quantitative model of phonetic convergencein spoken human-computer interaction and its application in asimulated spoken dialogue system environment and, specifically, itsspeech synthesis component. The project will build synthetic voicesfor statistical parametric speech synthesis that are capable ofadapting their speech output to the user's speech patterns, and deriveimplications for the design of conversational interfaces in speechtechnology. Subjective and objective evaluation methods will be usedto evaluate the success of the accommodation by the synthesizer to thehuman talker's speech characteristics and the acceptability of such acommunicative behavior displayed by the synthetic 'talker' from theperspective of the human interlocutor. A better understanding of theaccommodation phenomena related to various acoustic-prosodic,temporal, and spectral features may further improve the performance ofcurrent spoken dialogue system technology, leading to smootherconversational dialogues.

• DFG Programme: Research Grants
• Ehemaliger Antragsteller: Dr. Ingmar Steiner, until 12/2018