G2G: TTS-Driven Pronunciation Learning for Graphemic Hybrid ASR

TL;DR

This paper introduces a G2G model trained on TTS data to improve graphemic ASR, especially for rare words, achieving significant WER reductions without altering acoustic model training.

Contribution

The novel G2G approach enhances graphemic ASR by rewriting character sequences into more phonetic forms, improving recognition of rare words without changing existing acoustic models.

Findings

Word Error Rate reduced by 3% to 11%.

Bridges gap in rare name recognition.

No change needed in acoustic model training.

Abstract

Grapheme-based acoustic modeling has recently been shown to outperform phoneme-based approaches in both hybrid and end-to-end automatic speech recognition (ASR), even on non-phonemic languages like English. However, graphemic ASR still has problems with rare long-tail words that do not follow the standard spelling conventions seen in training, such as entity names. In this work, we present a novel method to train a statistical grapheme-to-grapheme (G2G) model on text-to-speech data that can rewrite an arbitrary character sequence into more phonetically consistent forms. We show that using G2G to provide alternative pronunciations during decoding reduces Word Error Rate by 3% to 11% relative over a strong graphemic baseline and bridges the gap on rare name recognition with an equivalent phonetic setup. Unlike many previously proposed methods, our method does not require any change to the acoustic model training procedure. This work reaffirms the efficacy of grapheme-based modeling and shows that specialized linguistic knowledge, when available, can be leveraged to improve graphemic ASR.