Hindi accent using deepspeech

How to train and evaluate on Hindi accent (speech to text).

  1. There is an audio file in Hindi mixed with English (few words that are used common) now i need translate Hindi audio to English as text.
  2. find the sentiment on the transcribed words

@bharath.vadithya I’m not sure I understand your post here, are you asking for how to do the three items ? Are you exploring alternatives to support Hindi ? Please elaborate.

Hi @lissyx… I have edited the question… i need those two items to complete. yes i am exploring alternatives to support Hindi, so that i an train my model better to understand Hindi and can transcribe to English. Thanks

Do you want to analyze emotions ? This is out of the scope of what we do

I’m unsure here, you want to transcribe Hindi, with a few common words of English ?

yeah i want to transcribe Hindi, with few common words of English.

alright, I understand this !!

Then you would need a training dataset in Hindi, I guess. And you could add the (expected?) english words in your language model?

yeah now i got the training data… where i can add it … (I am new to this area)

Can you read the documentation ? This should be covered quite correctly.

where it is ?? the gihub one !!

what about windows ??? actually i don’t have linux or mac

It may work but we don’t use that. It looks like TensorFlow GPU is supported there. That’s the best I can tell.

thanks :slight_smile: @lissyx

is there any ds_ctcdecoder for windows platform… the wheel file which is present in the github command line is linux supported.

No, we don’t have that (yet?).

but the above link says that if i can prefer … i can build the wheel file from source… here is the guide link iam following below

I am unsure what you are referring to here. We don’t support building the decoder for Windows, period.