A Platform for Rapid Development of Speech Recognition & Speech Synthesis Software for Under-Resourced Languages

点击翻译按钮获取中文摘要

"This project proposes a large-scale platform of automatic voice recognition (AVR) and text-to-speech (TTS) for under-resourced languages, i.e. languages for which data required to build statistical models of speech are limited, and difficult to acquire. AVR applies a mathematical model to convert speech into information a computer can understand and manipulate. TTS applies a statistical model to digital text to synthesize intelligible speech. Commercial AVR and TTS are available for highly-resourced languages of Europe, the Americas, developed Asia and China. Globally, speech products represent a £17b industry, and are found in voice-enabled devices such as Siri and Alexa, and in call centres for financial services, health care and telecoms companies. Many emerging market languages are under-resourced. Growing regional economies, along with product relevance to an expanding consumer class and rural communities establish a significant market opportunity.This project focuses on under-resourced African languages. Key challenges associated with developing AVR and TTS for African languages include modelling tonality, ligatures in colloquial speech, and unique phonemic patterns. This project addresses innovations related to the development of acoustic models for AVR & TTS for under-resourced languages."