Hi,
Yes, I include them in the average voice model training.
Sorry, I wasn't completely precise in my last letter. For average voice I
have (1800 declarative sentences + 300 questions) * 8 speaker. For
adaptation I have 100 declarative sentence + 100 questions. The modality
(/K:) context/question is included in the average voice model training
and
in the adaptation as well. I added two questions to the question file:
1. Is it a declarative sentence? (/K:0)
2. Is it a question? (/K:1)
What else should I pay attention to?
Thanks and regards,
Balint Toth
2013.05.01. 2:00 keltezéssel, Keiichiro Oura írta:
Hi,
Do you include /K: contexts/questions to average voice model training?
Regards,
Keiichiro Oura
2013/4/30 Tóth Bálint <toth.b@xxxxxxxxxxx>:
Dear All,
I would like to make HTS to ask questions. In the training database I
have 8
speakers, 1800 declarative sentence and 300 questions. I make
adaptation
with 200 sentences (100 declerative, 100 questions). I included in full
context labels an addition label regarding to the modality of the
sentence
(/K:0 = declarative, /K:1 = question). I included these features in the
question file as well.
After training and adaptation there are some changes in the intonation
in
case of questions, but it is very far from perfect.
What else should I do to make HTS to ask questions with good
intonation?
Thanks and regards,
Balint Toth