Vakyansh: ASR Toolkit for Low Resource Indic languages
TL
TL;DR
Vakyansh, an end to end toolkit for Speech Recognition in Indic languages introduces automatic data pipelines for data creation, model training, model evaluation and deployment and hopes that this will inspire the speech community to develop speech first applications using the ASR models inIndic languages.
Abstract
We present Vakyansh, an end to end toolkit for Speech Recognition in Indic languages. India is home to almost 121 languages and around 125 crore speakers. Yet most of the languages are low resource in terms of data and pretrained models. Through Vakyansh, we introduce automatic data pipelines for data creation, model training, model evaluation and deployment. We create 14,000 hours of speech data in 23 Indic languages and train wav2vec 2.0 based pretrained models. These pretrained models are then finetuned to create state of the art speech recognition models for 18 Indic languages which are followed by language models and punctuation restoration models. We open source all these resources with a mission that this will inspire the speech community to develop speech first applications using our ASR models in Indic languages.
References18 items
1
CLSRIL-23: Cross Lingual Speech Representations for Indic Languages
2
Automatic Speech Recognition in Sanskrit: A New Speech Corpus and Modelling Insights
3
Multilingual and code-switching ASR challenges for low resource Indian languages
4
iNLPSuite: Monolingual Corpora, Evaluation Benchmarks and Pre-trained Multilingual Language Models for Indian Languages
5