search

frank613 / CTC-based-GOP

This repo related to the paper "A Framework for Phoneme-Level Pronunciation Assessment Using CTC" for INTERSPEECH2024
13 stars 2 forks source link

I know all the words, but I don't know how to use them. #2

Open majisama opened 1 week ago

majisama commented 1 week ago

I know all the words, but I don't know how to use them.

majisama commented 1 week ago

Tell me directly how I can get a score of 0 to 1. God bless you!

majisama commented 1 week ago

I'm a layman, why does pronunciation assessment even require training asr

frank613 commented 1 week ago

Hello,

The original plan was we provide the script only for training the phoneme recogniserto reduce the size of our repository. In order to make it easily accessible , I uploaded a model that we fine-tuned for english. I also updated the README file and hope it helps.

BTW, the GOP is a score in the range [-inf, 0]. It is usually based on an ASR model to address the issue of data sparsity for pronunciation assessment.

Thanks, Xinwei

majisama commented 1 week ago

你好，

最初的计划是，我们只提供用于训练音素识别器的脚本，以减少 github 的大小。为了方便访问，我上传了一个我们针对英语进行了微调的模型。我还更新了 README 文件，希望对您有所帮助。

顺便说一句，GOP 是一个在 [-inf, 0] 范围内的分数。它通常基于 ASR 模型来解决发音评估的数据稀疏性问题。

谢谢， 新伟

Thank you. You're a good man. Taishanglaojun bless you.