The best results were obtained training an ASR system using a separate large corpus without pronunciation quality annotations and then adapting ...
確定! 回上一頁