The Seventh ISCA Tutorial and Research Workshop on Speech Synthesis
This study investigates the individual influences of the three prosodic cues response delay, the filler "hmm", and rising intonation on the perception of uncertainty of fictitious human-computer dialogue. Response delay is the time that the computer waits until it starts to answer a given question. the filler, i.e. the hesitation particls "hmm", can be insertes before the content of the answer starts. the final part of the answer's intonation contour can rise or fall. We hypothesize a hierarchy of influence: Rising intonation has a stronger influence on uncertainty perception than filler; response delay has the weakest effect. In a perception study the uncertainty of utterances generated with articulatory speech synthesis was tested. Results indicate that all cues have an effect on the perception of uncertainty, but the relative impact differs: Delay has a rather weak effect, whereas risinbg intonation and filler seem to be uncertainty-enhancing acoustic cues, each having strong effects which seem to override the weaker cue of delay. The results can serve as guideline for automatic detection of uncertainty in spoken dialogue systems.
Index Terms: uncertainty, prosody, paralinguistic expression, articulatory speech synthesis
Bibliographic reference. Lasarcyk, Eva / Wollermann, Charlotte (2010): "Do prosodic cues influence uncertainty perception in articulatory speech synthesis?", In SSW7-2010, 230-235.