With the wide-scale adoption of speech-based applications, the research community has started focusing on ways to improve naturalness and expressiveness of human-machine interactions. Among the directions pursued, the study of paralinguistic phenomena in conversation plays an important role. This project proposes to investigate conversational laughter, both from a fundamental research perspective as well as from an application viewpoint. The addressed research question concerns the context in which social laughter occurs, in particular the use of acoustic-prosodic cues in marking it. The findings of this investigation will directly feed into a spoken dialogue system, with the aim of increasing its perceived naturalness. In a first step, phrase- and turn-level analyses will be performed to identify the range of cues used by speakers to mark the use of laughter in conversation. Next, state-of-the-art signal processing methods combined with prosodic information will be used to automatically detect and segment laughter. This will allow the analysis of a larger number of conversational corpora which would, in turn, improve the generalizability of the results. Finally, a laughter-enhanced dialogue system will be implemented and its naturalness will be tested through perception experiments. It is expected that the laughter-enhanced system will be perceived more natural and that documented gender differences found in the production of laughter will be found also in its perception.
Fields of science
Call for proposal
See other projects for this call