Neural Grapheme-To-Phoneme Conversion with Pre-Trained Grapheme Models
Lu Dong, Zhiqiang Guo, Chao-Hong Tan, Ya-Jun Hu, Yuan Jiang, Zhen-Hua Ling
Abstract
Neural network models have achieved state-of-the-art performance on grapheme-to-phoneme (G2P) conversion. However, their performance relies on large-scale pronunciation dictionaries, which may not be available for a lot of languages. Inspired by the success of the pre-trained language model BERT, this paper proposes a pre-trained grapheme model called grapheme BERT (GBERT), which is built by self-supervised training on a large, language-specific word list with only grapheme information. Furthermore, two approaches are developed to incorporate GBERT into the state-of-the-art Transformer-based G2P model, i.e., fine-tuning GBERT or fusing GBERT into the Transformer model by attention. Experimental results on the Dutch, Serbo-Croatian, Bulgarian and Korean datasets of the SIGMORPHON 2021 G2P task confirm the effectiveness of our GBERT-based G2P models under both medium-resource and low-resource data conditions.
BibTeX
@inproceedings{icassp2022_neuralgraphemeto,
title = {Neural Grapheme-To-Phoneme Conversion with Pre-Trained Grapheme Models},
author = {Lu Dong and Zhiqiang Guo and Chao-Hong Tan and Ya-Jun Hu and Yuan Jiang and Zhen-Hua Ling},
booktitle = {ICASSP 2022},
year = {2022}
}