Skip to content

A Full Text-Dependent End to End Mispronunciation Detection and Diagnosis with Easy Data Augment Techniques

Notifications You must be signed in to change notification settings

cageyoko/CTC-Attention-Mispronunciation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 

Repository files navigation

“A Full Text-Dependent End to End Mispronunciation Detection and Diagnosis with Easy Data Augment Techniques“ https://arxiv.org/pdf/2104.08428.pdf

Implemented with TIMIT and L2-Arctic Database:

TIMIT : -
L2-Arctic: (https://psi.engr.tamu.edu/l2-arctic-corpus/)

Note:
CNN-RNN-CTC is baseline.
attention_aug is our best system.

Usage:

  1. Just need to change your kaldi_path in path.sh and your data_path in run.sh
  2. ./run.sh to get the decode sequence (decode_seq)
  3. mv decode_seq ./result/hyp ./mdd_result.sh

Next step: We will update the results on large datasets (such as Librispeech)

About

A Full Text-Dependent End to End Mispronunciation Detection and Diagnosis with Easy Data Augment Techniques

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published