msb-public
/
PaddleSpeech
mirror of https://github.com/PaddlePaddle/PaddleSpeech
1
0
Code Issues Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'eff363f695'
${ noResults }
PaddleSpeech/speechx/examples/ds2_ol/aishell
History
Yang Zhou 22c3f95057
fix fbank result 		2 years ago
..
local speedup ngram build 2 years ago
.gitignore update aishell rsl 2 years ago
README.md Update README.md 2 years ago
path.sh fix nnet output name 2 years ago
run.sh rm to_float32 flags, default is fbank 2 years ago
run_fbank.sh fix fbank result 2 years ago
utils refactor speechx egs 2 years ago

README.md

Aishell - Deepspeech2 Streaming

How to run

bash run.sh

Results

CTC Prefix Beam Search w/o LM

Overall -> 16.14 % N=104612 C=88190 S=16110 D=312 I=465
Mandarin -> 16.14 % N=104612 C=88190 S=16110 D=312 I=465
Other -> 0.00 % N=0 C=0 S=0 D=0 I=0

CTC Prefix Beam Search w/ LM

LM: zh_giga.no_cna_cmn.prune01244.klm

Overall -> 7.86 % N=104768 C=96865 S=7573 D=330 I=327
Mandarin -> 7.86 % N=104768 C=96865 S=7573 D=330 I=327
Other -> 0.00 % N=0 C=0 S=0 D=0 I=0

CTC WFST

LM: aishell train --acoustic_scale=1.2

Overall -> 11.14 % N=103017 C=93363 S=9583 D=71 I=1819
Mandarin -> 11.14 % N=103017 C=93363 S=9583 D=71 I=1818
Other -> 0.00 % N=0 C=0 S=0 D=0 I=1

LM: wenetspeech --acoustic_scale=1.5

Overall -> 10.93 % N=104765 C=93410 S=9780 D=1575 I=95
Mandarin -> 10.93 % N=104762 C=93410 S=9779 D=1573 I=95
Other -> 100.00 % N=3 C=0 S=1 D=2 I=0