what's where (short cuts)
(go to English Text version)
Top-level pages (text in Japanese):
Corpus-based speech synthesis
.
Intro - `recycling' and `indexing'
Personality in the voice
,
Index-based speech segment retrieval
,
which in turn calls:
Random-access Synthesis
,
Context-sensitive units
,
What's an index?
which in turn calls:
the philosophy behind chatr
,
labelling
, and
selection weights
Multilingual
synthesis
Making a
multi-speaker
database.
How our attitude towards
texts
has evolved.
What's coming
next
(calls
improving intonation
:
1
,
2
, and
3
)
Applications
.
Speech samples:
multilingual CHATR:
A few
English
voices
A few
Japanese
voices
Alan Alda
an American English voice
(see
Scientific American Frontiers
A
British English
voice (text and tts-synthesis)
A
Chinese
conversation (from the HKU CD-Rom)
A
Korean
conversation
Ohta san's Korean
Some
German
(from the Kiel Uni CD-Rom)
Kansai Japanese
(a joke - note the laughter)
multilingual Japanese
Ohira Toru san
(aka `the laughing salesman')
Chatr's children
Kuroyanagi Tetsuko:
Q&A
(how chatr works)
Tetsuko's samples
Tetsuko's opinion
Multilingual Tetsuko
Tetsuko's weather report
Korean Tetsuko
more data needed
The synthesis system:
The weather forecast
ATR nutalk
vs a
young female
version
Processing
multiple languages
in a text
Copy Synthesis
(no prediction involved)
How
focus
changes meaning
Why we don't use
signal processing
Feedback:
Let us hear your comments
Inquiries about Chatr
(C) Copyright ATR Interpreting Telecommunications Research Labs 1997
Acknowledgements
people who contributed to CHATR
This page prepared by
Nick Campbell
.