Train- ing joint-sequence based G2P require explicit grapheme- to-phoneme alignments which are not straightforward since graphemes and phonemes don't correspond ...
確定！ 回上一頁
