Samtalets prosodi/Prosody in conversation

Project leader

Project Details

Start date: 01/01/2010
End date: 31/12/2013


Riksbankens Jubileumsfond (RJ) granted the project Prosody in conversation (Samtalets prosodi) 5.2 MSEK for the years 2010-2012 (contract P09-0064:1-E). Applicant: Mattias Heldner.

The project investigated how people talking to each other jointly decide who should speak when, and the role of prosody in making these joint decisions. While prosody is by no means the only relevant information for this joint interaction control, we believe that it plays a central role (see e.g. Edlund & Heldner, 2005, and references mentioned therein). A detailed model of the prosody involved in interaction control is crucial both for producing appropriate conversational behaviour and for understanding human conversational behaviour. Both are required in order to reach our visionary goal, and represent the artificial conversational partner in the roles of speaker and listener in a conversation, respectively.

One line of inquiry within the project is the quantitative acoustic analysis of prosodic features in genuine spoken face-to-face conversations. The project focuses on local intonation patterns in the immediate vicinity of interactional events, such as transitions from (i) speech to pauses (within-speaker silences); (ii) speech to gaps (between-speaker silences, i.e. at speaker changes); and (iii) speech by one speaker to speech by another speaker (i.e. overlapping speech in speaker changes). In addition, we analyze selected interactional phenomena occurring on a longer time scale, such as pitch similarity across these interactional events and the overall tendency of interlocutors to grow increasingly similar as the conversation proceeds. This increasing interlocutor similarity reported in the literature under many names (e.g. entrainment, alignment, accommodation; see e.g. Edlund, Heldner, & Hirschberg, 2009 for an overview) has been reported for a great number of linguistic features, but we limit ourselves to prosody in this project.

In addition, the results of the acoustic analyses are fed into a second line of inquiry: studies of the effects of using or introducing such prosodic features in a conversation. These studies will include listening experiments where manipulations of genuine conversations by means of re-synthesis are used as stimuli. Furthermore, there will be pragmatic experiments where the conversational behaviour in response to the use of such prosodic features in artificial speech is analyzed. Finally, there will be analyses of conversational behaviour in response to real-time (or minimum delay) manipulations of genuine conversations, such as deletions, insertions or manipulations of features.

External Partners

Last updated on 2017-24-03 at 12:39