Adversarial Teacher-Student Learning for Unsupervised Domain Adaptation

TL;DR

This paper introduces adversarial teacher-student learning for unsupervised domain adaptation in speech recognition, explicitly learning condition-invariant features to improve robustness against variability and noise.

Contribution

It proposes a novel adversarial T/S learning framework that jointly optimizes for condition invariance and robustness, extending prior T/S methods with adversarial training and multi-factorial adaptation.

Findings

Achieved 44.60% relative WER reduction over the source model.

Achieved 5.38% relative WER reduction over baseline T/S learning.

Demonstrated effectiveness on noisy CHiME-3 dataset.

Abstract

The teacher-student (T/S) learning has been shown effective in unsupervised domain adaptation [1]. It is a form of transfer learning, not in terms of the transfer of recognition decisions, but the knowledge of posteriori probabilities in the source domain as evaluated by the teacher model. It learns to handle the speaker and environment variability inherent in and restricted to the speech signal in the target domain without proactively addressing the robustness to other likely conditions. Performance degradation may thus ensue. In this work, we advance T/S learning by proposing adversarial T/S learning to explicitly achieve condition-robust unsupervised domain adaptation. In this method, a student acoustic model and a condition classifier are jointly optimized to minimize the Kullback-Leibler divergence between the output distributions of the teacher and student models, and simultaneously, to min-maximize the condition classification loss. A condition-invariant deep feature is learned in the adapted student model through this procedure. We further propose multi-factorial adversarial T/S learning which suppresses condition variabilities caused by multiple factors simultaneously. Evaluated with the noisy CHiME-3 test set, the proposed methods achieve relative word error rate improvements of 44.60% and 5.38%, respectively, over a clean source model and a strong T/S learning baseline model.