wav2vec2-mms-1b-cmn-phonetic

国语读音 ASR,将语音转换为读音以进行自动 TTS 标注

  • CER 0.022529

输出格式

使用一种修改版的注音实现发音完全对应, 可以转换回注音/拼音

initials = [
    "ㄅ",
    "ㄆ",
    "ㄇ",
    "ㄈ",
    "ㄉ",
    "ㄊ",
    "ㄋ",
    "ㄌ",
    "ㄍ",
    "ㄎ",
    "ㄏ",
    "ㄐ",
    "ㄑ",
    "ㄒ",
    "ㄓ",
    "ㄔ",
    "ㄕ",
    "ㄖ",
    "ㄗ",
    "ㄘ",
    "ㄙ",
    # IPA /j/ sound, when initial is absent while finals start with "ㄧ" or "ㄩ"
    # such as ㄧㄚ /jiɑ/ or ㄧㄝ /jiɛ/ or ㄩㄝ /jyɛ/
    "j",
]

finals = [
    "ㄚ",
    "ㄛ",
    "ㄜ",
    "ㄝ",
    "ㄞ",
    "ㄟ",
    "ㄠ",
    "ㄡ",
    "ㄢ",
    "ㄣ",
    "ㄤ",
    "ㄥ",
    "ㄦ",
    # special medials that can be used as prefixes for other medials
    "ㄧ",
    "ㄧㄛ",
    "ㄧㄚ",
    "ㄧㄝ",
    "ㄧㄠ",
    "ㄧㄡ",
    "ㄧㄢ",
    "ㄧㄣ",
    "ㄧㄤ",
    "ㄧㄥ",
    "ㄨ",
    "ㄨㄚ",
    "ㄨㄛ",
    "ㄨㄞ",
    "ㄨㄟ",
    "ㄨㄢ",
    "ㄨㄣ",
    "ㄨㄤ",
    "ㄨㄥ",
    "ㄩ",
    "ㄩㄝ",
    "ㄩㄢ",
    "ㄩㄣ",
    "ㄩㄥ",
    # https://zh.wikipedia.org/zh-tw/%E7%A9%BA%E9%9F%BB
    "ㄭ+",  # after ㄓ, ㄔ, ㄕ, ㄖ
    "ㄭ-",  # after ㄗ, ㄘ, ㄙ
    # https://zh.wikipedia.org/wiki/%E5%85%92%E5%8C%96
    "r",
]
Downloads last month
5
Safetensors
Model size
1.0B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Chuatury/wav2vec2-mms-1b-cmn-phonetic

Finetuned
(374)
this model

Dataset used to train Chuatury/wav2vec2-mms-1b-cmn-phonetic