G2G: TTS-Driven Pronunciation Learning for Graphemic Hybrid ASR
This work addresses a specific bottleneck in grapheme-based ASR for handling rare words, offering an incremental improvement that leverages linguistic knowledge.
The paper tackles the problem of rare long-tail words in grapheme-based automatic speech recognition (ASR) by training a statistical grapheme-to-grapheme model on text-to-speech data to rewrite character sequences into phonetically consistent forms. This method reduces Word Error Rate by 3% to 11% relative over a baseline and improves rare name recognition without modifying acoustic model training.
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.