troublem1 / wolof_asr-zindi-challenge

Rank 11/47

AI4D Baamtu Datamation - Automatic Speech Recognition in WOLOF

Can you create an automatic speech recognition model for Wolof for use in public transport?

BRIEF DESCRIPION

In a country such as Senegal, where about 50% of the population is illiterate, technologies and applications that are designed to be used by people who can read are not as effective as they could be. In this competition, your aim is to use Automatic Speech Recognition (ASR) techniques in the Wolof language to help illiterate people to interact with apps with just their voice, in a language they can already speak.

The goal of this competition is to build an ASR model that will help illiterate people use existing apps to find which bus they can take to reach their destination, without having to know how to read or write. check it out on Zindi

Environment

• Conda Python 3.7.9
• cuda Tesla P100
• Kaggle

input data should be structured as follows

How to run the code

having the data in te input directory, run Bamtu full ds.ipynb to create for new directory containing varients of the dataset.

Run Bamtu Full ds json.ipynb to generate manifet files train and test for sythetic dataset. Run in order

• fork_of_prep_files_before_major_training_on_00_asr .ipynb to output trainedmodel2_v9.nemo
• run 10-retrain-fork-prepfilesbeforemajortraining.ipynb to output forktrainedmodel10_v1.nemo
• 10-1-reretrain-fork-prepfilesbeforemajortraining.ipynb to output forktrainedmodel11_v1.nemo
• 10-1-1-reretrain-fork-prepfilesbeforemajortraining .ipynb to output forktrainedmodel111_v1.nemo
• 10-1-1-5reretrain-fork-prepfilesbeforemajortrain.ipynb to output forktrainedmodel1115_v1.nemo

wolof_asr-zindi-challenge's Issues

Add Requirements.txt

add kaggle json manifests and notebook Configs

Prepare notebooks

Create and Commit model configs