Hi everybody,

I'm trying to create an simple and small acoustic model for sphinx3 using SphinxTrain.

The tutorial shipped with SphinxTrain works fine and i get no errors. But when I create my own dictionary, phone and transcription files the verify script tells me that ALL the phones from the phone file cannot be found in the transcription file, for example this is the message for the phone P:

WARNING: This phone (P) occurs in the phonelist (xyz/testam/etc/testam.phone), but not in any word in the transcription (/xyz/testam/etc/testam_train.transcription)

I already read the other thread dealing with this issue, but the solution (converting CR/LF into unix style LF) does not work - i create all files using linux. I also tried creating the transcription and dictionary using the online lmtool provided by CMU, but it's all the same.

Maybe it's important to say that I have only one word in each utterance?

Well, maybe someone can help me... I have no idea why I get these errors. I uploaded the phone, dictionary and transcription files:

http://www.2shared.com/file/11682825/3b482ebd/testam.html

http://www.2shared.com/file/11682873/af5c7fcd/testam.html

http://www.2shared.com/file/11682944/1bd7d39a/testam_train.html?

Thank you in advance!

--- (Edited on 2/26/2010 5:41 am [GMT-0600] by Visitor) ---

> I'm trying to create an simple and small acoustic model for sphinx3 using SphinxTrain.

Not sure how this is related to Voxforge and why are you asking it here. You'll get faster answer on CMUSphinx forums.

> me that ALL the phones from the phone file cannot be found in the transcription file, for example this is the message for the phone P:

Your transcription file has incorrect format. It doesn't have fileid for each transcription in braces as a last word in the line. See an4 transcription format for example.

--- (Edited on 2/26/2010 20:48 [GMT+0300] by nsh) ---