Minimum Word Error Rate Training with Language Model Fusion for End-to-End Speech Recognition

TL;DR

This paper introduces a novel minimum WER training method with internal LM estimation fusion for end-to-end speech recognition, reducing the need for LM weight tuning and improving WER across multiple test sets.

Contribution

It proposes MWER-ILME, a new training approach that integrates internal LM estimation into MWER training, enhancing domain robustness without extensive LM weight tuning.

Findings

MWER-ILME achieves 8.8% relative WER reduction over MWER.

MWER-ILME outperforms MWER-SF with 5.8% WER reduction.

Robust LM integration across diverse test sets.

Abstract

Integrating external language models (LMs) into end-to-end (E2E) models remains a challenging task for domain-adaptive speech recognition. Recently, internal language model estimation (ILME)-based LM fusion has shown significant word error rate (WER) reduction from Shallow Fusion by subtracting a weighted internal LM score from an interpolation of E2E model and external LM scores during beam search. However, on different test sets, the optimal LM interpolation weights vary over a wide range and have to be tuned extensively on well-matched validation sets. In this work, we perform LM fusion in the minimum WER (MWER) training of an E2E model to obviate the need for LM weights tuning during inference. Besides MWER training with Shallow Fusion (MWER-SF), we propose a novel MWER training with ILME (MWER-ILME) where the ILME-based fusion is conducted to generate N-best hypotheses and their posteriors. Additional gradient is induced when internal LM is engaged in MWER-ILME loss computation. During inference, LM weights pre-determined in MWER training enable robust LM integrations on test sets from different domains. Experimented with 30K-hour trained transformer transducers, MWER-ILME achieves on average 8.8% and 5.8% relative WER reductions from MWER and MWER-SF training, respectively, on 6 different test sets