• Wentao Gu

    Nanjing Normal University, China

    • 11:00:15 - 17-08-2020

      Prosody-oriented studies of social communicative speech in a digitized world

      In the last decade, human-human and human-machine social communicative dialogues have received more and more attention among linguists and speech scientists. For one thing, linguists study social communicative speech in order to go beyond controlled experiments and get additional insights into how spoken languages are processed in dynamic interaction. For another thing, accurate and phenomenologically rich automatic speech/speaker/emotion recognition and expressive text-to-speech synthesis systems are essential for conversational dialogue systems, as these become increasingly more interactional and social rather than solely transactional. Prosody, as the major vehicle of social functions, plays key roles in both types of studies. The investigation of prosodic variation in dialogue does not only require applying existing methods to interactional data. It also requires developing new categories of forms and functions, new modeling techniques and new sources of data/knowledge. This special session aims at bringing together phoneticians, linguists and speech technologists interested in the prosody of conversational speech. Submissions on the following topics and on different languages, including minority languages, are especially welcome:
      • tools and data resources for the annotation and analysis of prosody in conversational speech
      • the relationship between prosodic forms and communicative functions
      • cross-linguistic and individual prosody variation in social speech communication
      • co-variation of the segmental and suprasegmental characteristics in speech communication
      • models of prosody in conversational speech
      • prosody-oriented studies in automatic speech/speaker/emotion recognition and expressive text-to-speech synthesis
      • prosody-oriented studies in human-robot interaction

      The special session is able to include a maximum of 10 submissions, e.g., subdivided into 7 oral presentations of 15 minutes plus 5 minutes for discussion and 3 oral presentations that showcase new methods, tools, and corpora. Submissions to this special session of TAI use the same abstract and paper templates as regular contributions to TAI, and submissions are made through the same Easychair link as for regular contributions to TAI (see the TAI website, call for papers).