GGitHub【Hackathon 8th No.7】Python版本适配 3 (#3969 )

85de840d创建于 2025年1月21日历史提交

文件	最后提交记录	最后更新时间
README.md	Added pre-install doc for G2P and TN modules and updated the dependency version of pypinyin (#2364) * Added pre-install doc for G2P and TN modules and updated the dependency version of pypinyin, test=doc	3 年前
compare_badcase.py	add tool to compare test badcase and add run examples,test=tts	3 年前
get_g2p_data.py	【Hackathon 8th No.7】Python版本适配 3 (#3969) * Update setup.py * add optional * fit with praatio>=6.0.0 * Apply suggestions from code review * Apply suggestions from code review * Apply suggestions from code review	1 年前
path.sh	[TTS]fix praatio version, test=tts (#1158) * fix praatio version, test=tts * fix praatio version, test=tts	4 年前
run.sh	[Fix] g2p run.sh (#3886)	1 年前
test_g2p.py	change nprocs to ngpu, add aishell3/voc1	4 年前

G2P

For g2p, we use BZNSYP's phone label as the ground truth and we delete silence tokens in labels and predicted phones.

You should Download BZNSYP from its Official Website and extract it. Assume the path to the dataset is ~/datasets/BZNSYP.

We use WER as an evaluation criterion.

Start

Run the command below to get the results of the test.

cd ../../../tools
bash extras/install_sclite.sh
cd -
./run.sh

The avg WER of g2p is: 0.024075726733983775

     ,--------------------------------------------------------------------.
     |                         ./exp/g2p/text.g2p                         |
     |--------------------------------------------------------------------|
     | SPKR   | # Snt    # Wrd  | Corr    Sub    Del    Ins    Err  S.Err |
     | Sum/Avg|  9996   299181  | 97.6    2.4    0.0    0.0    2.4   49.0 |
     `--------------------------------------------------------------------'