Sequence Training of DNN Acoustic Models With Natural Gradient

TL;DR

This paper introduces a Natural Gradient-based sequence training method for DNN acoustic models, improving convergence speed and applicability over traditional SGD and Hessian Free methods, with demonstrated efficiency and accuracy gains.

Contribution

It presents a novel Natural Gradient optimization framework for sequence training of DNN acoustic models, applicable to various criteria and outperforming existing methods in convergence speed.

Findings

Faster convergence compared to Hessian Free training.

Effective application across different sequence discriminative criteria.

Improved transcription accuracy on MGB task.

Abstract

Deep Neural Network (DNN) acoustic models often use discriminative sequence training that optimises an objective function that better approximates the word error rate (WER) than frame-based training. Sequence training is normally implemented using Stochastic Gradient Descent (SGD) or Hessian Free (HF) training. This paper proposes an alternative batch style optimisation framework that employs a Natural Gradient (NG) approach to traverse through the parameter space. By correcting the gradient according to the local curvature of the KL-divergence, the NG optimisation process converges more quickly than HF. Furthermore, the proposed NG approach can be applied to any sequence discriminative training criterion. The efficacy of the NG method is shown using experiments on a Multi-Genre Broadcast (MGB) transcription task that demonstrates both the computational efficiency and the accuracy of the resulting DNN models.