Advance research in agricultural text-to-speech: the word segmentation of analytic language and the deep learning-based end-to-end system

Abstract

Agricultural Text-to-Speech (TTS) has attracted increasingly more attention. The application of agricultural TTS and its problems are analyzed in this paper, and the traditional framework of the TTS system and its key technologies, i.e., text analysis, rhythm generation and speech synthesis are discussed. Furthermore, two advancements in agricultural TTS, the word segmentation of analytic language and the deep learning-based end-to-end TTS system, are detailed summarized. Based on the characteristics of agriculture, some appealing research directions are pointed out: how to improve the training speed and synthesis speed of the deep learning models is still the focus; the study on the approaches of weakly-supervised learning in TTS is in fancy; and research on the real-time and high-quality speech synthesis that can be deployed in mobile devices is a key point of agricultural TTS research.