Information Technology Reference
In-Depth Information
linguistic analysis of the words describing hesitation related phenomena and to cluster
their occurrence according to their semantic similarity and the context in which they
occur. This will be future work.
3.3 Speech Analysis
Hesitation marking in speech is fairly varied. In English this ranges from different
types of hesitation markers (
uhh, umm
) to pauses (silence) and slow speaking rate.
Fundamental frequency F0 is also shown to rise before pauses occurring in major
syntactic boundaries, but not if the pause occurs elsewhere. In Japanese, [10] found
that the prosodic and temporal features of a response carry information about how the
speaker has grounded the information expressed in the partner's previous utterance,
especially if the speaker repeats a portion of that utterance. The features such as
longer delays, higher pitch, slower tempo, and rising boundary tone signal lower
integration degree, i.e. the information is not fully integrated in the speaker's body of
knowledge, and in the context of this paper we can say that they express hesitation
and uncertainty.
Figure 3 (a screenshot by the Praat software, http://www.fon.hum.uva.nl/praat/)
shows speech analysis of the end part of the speaker's utterance
it's just a…
which
occurs immediately before a longer silence and the shoulder shrug. In the screenshot,
red points show variation of the first format F0, the blue line shows the pitch contour,
and the green continuous line shows intensity.
Fig. 3.
Combined speech analysis in Praat (see text)
It can be seen that the utterance conforms to all the typical prosodic characteristics
related to hesitations, and does not differ from the speaker's other hesitative
utterances either. The speaker does not raise his voice (mean energy is about 60 dB),
but lengthens the function word so that the speaking rate seems to slow down. The
pitch contour stays flat, and shows no upward tendency immediately before the pause
(and the shoulder shrug), which is expected since the pause does not occur at a major
syntactic boundary.
Considering the correlation of prosody and hesitation related communicative body
movements, an interesting question is thus if the two modalities, speech and gesturing
support each other or if they operate independently of in the communicative situation.
It seems that the shoulder shrugging gesture and prosodic marking have no special
correlation in this particular case: the shrug occurs during the pause and it seems more
natural to relate the prosody to the unfinished utterance and uncertainty in general,
