edaiofficial / okwugbe Goto Github PK

In a talk with Spectie, he told me about https://github.com/ftyers/commonvoice-utils

It provides two things:

1. a better way of getting authentic alphabets from CV languages
   so as i understand i only need to run this to get the alphabet for the language:

from cvutils import Alphabet
a = Alphabet('cv')
a.get_alphabet()
' -абвгдежзийклмнопрстуфхцчшщыэюяёҫӑӗӳ' 

2. a way to clean the transcripts (remove punctuation marks, non-alphabet characters, etc

I will work on this ASAP.

Let's say I have a train.csv file but the audio path in my csv file is not the correct path of the audio (possible reasons are generating your csv file on the local PC and having to upload the audio files to Colab to use).

We need the user to be able to specify audio_path while calling TrainOkwuGbe