[hts-users:02467] Re: problem building regression-class trees and adaptation problem with xform mask
- Subject: [hts-users:02467] Re: problem building regression-class trees and adaptation problem with xform mask
- From: Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>
- Date: Mon, 10 May 2010 20:57:10 +0900
- Cc: uratec <uratec@xxxxxxxxxxxxxxx>
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:sender:received :in-reply-to:references:date:x-google-sender-auth:message-id:subject :from:to:cc:content-type:content-transfer-encoding; bh=YrQbqoe5KuHUf28yQ/qQoK7cqsE/jqrl7YStpfoaIFs=; b=Cu9tII4upfdqlxwSqbbFsvLgQSTl6AgXIRXoSsq0GrvlR8Fbk5QvajCbSp9izFWVmf e9Y8DZW/+e1f/zccD8LF4n6Gc9+svYj5bv9/q4irRQgvidGN5L9sC6sZcWrTJuQhm3P5 SSAwQdbRikCWIZpuvEFIwwoNkMQWcDnoCr3nU=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:date :x-google-sender-auth:message-id:subject:from:to:cc:content-type :content-transfer-encoding; b=ddec6vzYv4IqZiXbbv0mE8pvze/SawtNu93rYM8a5k+OdGHJEPCrtlUVL1884ekeL3 EvPD7xHteRTOSZUILt3d9wdbTBm0Wxc3cOV6dWW/yeWR2TCcYn7k/7XU7t76dYRXMot3 OaF7NVyjaNlheNUxWj7DNh6mpvPXD/oHfdqUE=
Hi,
>>First error:
>>How large database do you use?
>>Does your machine have enough memory?
>
> The database is 1000 sentences. The database for the adaptation is 40
> sentences. I have used same database (1000 sent.) and even larger (3000
> sent.) for the demo without the adaptation and there is no problem. The
> machine has 3 GB ram. Is not enough?
Your machine have enough memory.
It can be seen that your scirpts and HTS-demo script are not the same.
If you use HTS-demo scirpts, does the same problem occur?
>>Second error:
>>Please, set appropriate speaker name pattern to -h option.
>>If filenames are Mike_001.cmp and John_001.cmp, */%%%%_* should be set.
>
> Do you mean that the files which are used for training (e.g. Mike) and the
> files which are used for the adaptation (e.g. John) should be of the same
> number of letter?
Speaker names in the training files must consist of the same number of letter.
Adaptation data is not used in SAT step.
In my examples, both `Mike_001.cmp' and `John_001.cmp' are training data.
Regards,
Keiichiro Oura
>
> Kind regards,
> Alexandros Lazaridis
>
>
> ========================================
> Alexandros Lazaridis
> Electrical and Computer Engineer
> PhD Student
>
> Artificial Intelligence Group
> Wire Communications Laboratory
> Department of Electrical & Computer Engineering
> University of Patras
> Patras 265 00, Greece
>
> Email: alaza@xxxxxxxxxx
> Tel: +30 2610 996496
> Fax: +30 2610 997336
> http://www.wcl.ece.upatras.gr/ai/
> http://www.wcl.ece.upatras.gr/alaza/
> BRING THEM BACK
> http://www.bringthemback.org/
> ========================================
>
>
>
>
>
- References
-
- [hts-users:02461] Re: problem building regression-class trees and adaptation problem with xform mask, Keiichiro Oura
- [hts-users:02462] Re: problem building regression-class trees and adaptation problem with xform mask, Alexandros Lazaridis