arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.10716cs.CLeess.AS
收藏

基于预先训练的字素模型的神经网络字素-音素转换

Neural Grapheme-to-Phoneme Conversion with Pre-trained Grapheme Models

iFLYTEK Research(科大讯飞研究院)

Lu Dong, Zhi-Qiang Guo, Chao-Hong Tan, Ya-Jun Hu, Yuan Jiang, Zhen-Hua Ling

摘要:

神经网络模型在字素到音素(G2 P)转换方面已经取得了最先进的性能。然而,它们的性能依赖于大规模的发音字典,这可能不适用于许多语言。受预训练语言模型BERT成功的启发,本文提出了一种称为字素BERT(GBERT)的预训练字素模型,该模型是通过对一个仅包含字素信息的大型特定语言单词列表进行自监督训练而构建的。此外,开发了两种方法来将GBERT纳入最先进的基于Transformer的G2 P模型,即,微调GBERT或通过注意力将GBERT融合到Transformer模型中。在SIGMORPHON 2021 G2 P任务的荷兰、塞尔维亚-克罗地亚、保加利亚和韩国数据集上的实验结果证实了我们基于GBERT的G2 P模型在中等资源和低资源数据条件下的有效性。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑