End-to-End Bengali Speech Recognition

TL;DR

This paper develops an end-to-end Bengali speech recognition system using CNN-RNN models with novel small convolution kernels, achieving competitive word error rates and demonstrating the effectiveness of efficient CNN blocks in low-resource language ASR.

Contribution

It introduces and evaluates small 7x3 and 3x3 convolution kernels within CNN-RNN architectures for Bengali ASR, showing improved performance over larger kernels.

Findings

Best model achieves 13.67% WER

Small kernels reduce FLOPs and parameters

Significant WER reduction over larger kernels

Abstract

Bengali is a prominent language of the Indian subcontinent. However, while many state-of-the-art acoustic models exist for prominent languages spoken in the region, research and resources for Bengali are few and far between. In this work, we apply CTC based CNN-RNN networks, a prominent deep learning based end-to-end automatic speech recognition technique, to the Bengali ASR task. We also propose and evaluate the applicability and efficacy of small 7x3 and 3x3 convolution kernels which are prominently used in the computer vision domain primarily because of their FLOPs and parameter efficient nature. We propose two CNN blocks, 2-layer Block A and 4-layer Block B, with the first layer comprising of 7x3 kernel and the subsequent layers comprising solely of 3x3 kernels. Using the publicly available Large Bengali ASR Training data set, we benchmark and evaluate the performance of seven deep neural network configurations of varying complexities and depth on the Bengali ASR task. Our best model, with Block B, has a WER of 13.67, having an absolute reduction of 1.39% over comparable model with larger convolution kernels of size 41x11 and 21x11.