Work

Adversarial Text-to-Speech for low-resource languages

Year: 2022

Type: conference-paper

Abstract: In this paper we propose a new method for training adversarial text-to-speech (TTS) models for low-resource languages using auxiliary data.Specifically, we modify the MelGAN (Kumar et al., 2019) architecture to achieve better performance in Arabic speech generation, exploring multiple additional datasets and architectural choices, which involved extra discriminators designed to exploit high-frequency similarities between languages.In our evaluation, we used subjective human evaluation, MOS -Mean Opinion Score, and a novel quantitative metric, the Fréchet Wav2Vec Distance, which we found to be well correlated with MOS.Both subjectively and quantitatively, our method outperformed the standard MelGAN model.

Source: Arabic Natural Language Processing Workshop (WANLP)

DOI: 10.18653/v1/2022.wanlp-1.8

Authors: Ashraf Elneima, Mikołaj Bińkowski

Institutions: African Institute for Mathematical Sciences, African Institute for Mathematical Sciences

Language: English

Topic: Speech Recognition and Synthesis

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

OpenAlex ID: W4385574089

FWCI: 0.829

Cites: 22

Cited by: 6

Related to: 10