Hi Hiyassat,
Please drop me a mail once you have established the Arabic script. In the mean time, if you have any questions on Sphinx and SphinxTrain. Just give me a mail.
Arthur
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Are you read the documentation for the SphinxTrain?
If your answer is yes, then you can use the SphinxTrain scripts (perl scripts) for Train a aucustic model, this way is simple ut you need understand the concepts an arguments that the SphinxTrain needed.
What OS are you using for the Training? if is Linux then you have many programs needed intalled now.
Now the first step in the training is read the documentation for the Sphinx Train and unserstand
In this forum exist many busy people that can help you, Only ask the right question, and please give ours many information about your questions.
Omar
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
my os is linux
i tried the SphinxTrain but i failed because this based on the english dictionary and i need arabic recognition
also i have proplem in makedict script
it ask about fistival but i havent any clue about fistival
thank u
hussein
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
first of all, you need to read and understand the documentation about SphinxTrain. It is important to know what are you exactly doing in each of the steps of the proccess in order to try to solve the difficulties of the training.
Makedict only work for the english language. You need a Grapheme-to-phoneme conversion for your language, i.e., you need a translator from a word to the pronunciation of this word. I have trained several acoustic models for the spanish language with SphinxTrain and we have written a grapheme-to-phoneme translator for the castilian-spanish. I don't know if there are a "makedict" for the arabic language. If it doesn't exist you need to create one in order to use the SphinxTrain utilities.
I hope this helps.
Sergio
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2005-06-22
A grapheme-to-phone translator for the new language would be useful, but perhaps too difficult and not necessary in all cases.
What is actually needed is a dictionary file containing all the words in the training database and their pronunciations. If the training set is not too large, simply create this data manually.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
to be the standard place for SphinxTrain documentation.
Please also accept my apology on the misleading message in tinydoc.txt. That document is supposed to be used by developer who want to train an English model only. I have already file a bug request and is planning to fix it in later release myself.
Arthur
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Arthur
Thank you for your support
I al ready develop an automatic tool to convert Arabic grapheme to phoneme i.e. to create dictionary once it is ready I will make it available for you and any one interested.
also I train my model using this dictionary and obtain very good result
also I have three gigas corpus for Arabic I will make it available as soon as I establish my own web site.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
i am new user to sphinx and i try to build acoustic model for arabic language colud some body help me how to start and what should i do
Hi Hiyassat,
Please drop me a mail once you have established the Arabic script. In the mean time, if you have any questions on Sphinx and SphinxTrain. Just give me a mail.
Arthur
Hi,
Before that,
Are you read the documentation for the SphinxTrain?
If your answer is yes, then you can use the SphinxTrain scripts (perl scripts) for Train a aucustic model, this way is simple ut you need understand the concepts an arguments that the SphinxTrain needed.
What OS are you using for the Training? if is Linux then you have many programs needed intalled now.
Now the first step in the training is read the documentation for the Sphinx Train and unserstand
In this forum exist many busy people that can help you, Only ask the right question, and please give ours many information about your questions.
Omar
my os is linux
i tried the SphinxTrain but i failed because this based on the english dictionary and i need arabic recognition
also i have proplem in makedict script
it ask about fistival but i havent any clue about fistival
thank u
hussein
Hi,
first of all, you need to read and understand the documentation about SphinxTrain. It is important to know what are you exactly doing in each of the steps of the proccess in order to try to solve the difficulties of the training.
Makedict only work for the english language. You need a Grapheme-to-phoneme conversion for your language, i.e., you need a translator from a word to the pronunciation of this word. I have trained several acoustic models for the spanish language with SphinxTrain and we have written a grapheme-to-phoneme translator for the castilian-spanish. I don't know if there are a "makedict" for the arabic language. If it doesn't exist you need to create one in order to use the SphinxTrain utilities.
I hope this helps.
Sergio
A grapheme-to-phone translator for the new language would be useful, but perhaps too difficult and not necessary in all cases.
What is actually needed is a dictionary file containing all the words in the training database and their pronunciations. If the training set is not too large, simply create this data manually.
Omar and Sergio,
Thanks for answering this mail.
Just to supplement your answer, you will find
http://fife.speech.cs.cmu.edu/sphinxman/fr4.html
to be the standard place for SphinxTrain documentation.
Please also accept my apology on the misleading message in tinydoc.txt. That document is supposed to be used by developer who want to train an English model only. I have already file a bug request and is planning to fix it in later release myself.
Arthur
Arthur
Thank you for your support
I al ready develop an automatic tool to convert Arabic grapheme to phoneme i.e. to create dictionary once it is ready I will make it available for you and any one interested.
also I train my model using this dictionary and obtain very good result
also I have three gigas corpus for Arabic I will make it available as soon as I establish my own web site.