CLSRIL-23: Cross Lingual Speech Representations for Indic Languages

TL;DR

CLSRIL-23 is a self-supervised, multilingual speech model trained on 23 Indic languages that improves speech recognition performance, especially for low-resource languages, by learning shared phonetic representations.

Contribution

The paper introduces CLSRIL-23, a novel self-supervised model for cross-lingual speech representation learning across 23 Indic languages, leveraging multilingual pretraining to enhance downstream speech recognition tasks.

Findings

Multilingual pretraining outperforms monolingual training.

5% reduction in WER for Hindi speech recognition.

9.5% reduction in CER for Hindi speech recognition.

Abstract

We present a CLSRIL-23, a self supervised learning based audio pre-trained model which learns cross lingual speech representations from raw audio across 23 Indic languages. It is built on top of wav2vec 2.0 which is solved by training a contrastive task over masked latent speech representations and jointly learns the quantization of latents shared across all languages. We compare the language wise loss during pretraining to compare effects of monolingual and multilingual pretraining. Performance on some downstream fine-tuning tasks for speech recognition is also compared and our experiments show that multilingual pretraining outperforms monolingual training, in terms of learning speech representations which encodes phonetic similarity of languages and also in terms of performance on down stream tasks. A decrease of 5% is observed in WER and 9.5% in CER when a multilingual pretrained model is used for finetuning in Hindi. All the code models are also open sourced. CLSRIL-23 is a model trained on $23$ languages and almost 10,000 hours of audio data to facilitate research in speech recognition for Indic languages. We hope that new state of the art systems will be created using the self supervised approach, especially for low resources Indic languages.