The MIL Speech Seminars occur roughly every two weeks during full term and occasionally during vacations. The purpose of the seminars is to enable researchers in various fields to present recent work that has potential benefits for speech technology.
The MIL Speech Seminar series schedule for Michaelmas Term was:
|22nd October 2007||Francois Mairesse (University of Sheffield)||Generating Language with Personality||
Over the last fifty years, the "Big Five"
model of personality traits has become a
standard in psychology, and research has
systematically documented correlations between
a wide range of linguistic variables
and the Big Five traits. A distinct line of
research has explored methods for automatically
generating language that varies along
We present PERSONAGE (PERSONAlity GEnerator), the
first highly parametrisable language generator
for all Big Five personality traits. We evaluate
two personality generation methods: (1) direct
generation with particular parameter settings
suggested by the psychology literature; and
(2) overgeneration and selection using statistical
models trained from judge's ratings.
Results show that both methods reliably generate
utterances that vary along the extraversion
dimension, according to human judges. We also
report results for other personality dimensions,
with the aim to produce language expressing any combination of personality traits.
|12th November 2007||Sacha Krstulovic (Toshiba)||HMM synthesis developments at Toshiba||
Recently, HMM-based speech synthesis has been drawing lots of
attention, not only due to its competitive and stable quality,
but also and mainly due to its flexibility. As a matter of fact,
it is based on one of the best known and most complete parametric
models of speech, which, once trained, allows to perform some
alterations of the synthetic voice independently of the original
After a presentation of the Toshiba company, its speech research
and more specifically its research interests in Text-To-Speech
synthesis, this talk will present an overview of the use of HMMs
for speech synthesis, while underlining some contrasts with their use
in Automatic Speech Recognition. Some audio examples of HMM synthesized
sentences will be played. The talk will conclude with some current
and future research directions undertaken by Toshiba in this field.
If you are interested in giving a seminar presentation or if you would like more information about the seminar series please contact Marcus Tomalin