Deleted Interpolation model is giving more error rate than CD and CI model

Speech Recognition Toolkit

Brought to you by: air, arthchan2003, awb, bhiksha, and 5 others

This project can now be found here.

Deleted Interpolation model is giving more error rate than CD and CI model

Forum: Help

Creator: puneet

Created: 2018-02-09

Updated: 2018-02-09

puneet - 2018-02-09

Hi every one
I trained phonetically tied acoustic model (PTM)---4 models were generated-- CI, CD, CD_Untied and CD_DelInterpolation. The CD_DelInterpolation model showed highest Word error rate than all other models. I studied literature which indicated that Dleted interpolation results in lower WER. Please some one help me out. I have used Pocketsphinx decoder for it.
Thanks

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- Nickolay V. Shmyrev - 2018-02-09
  
  It depends a lot on the dataset, overall, we do not recommend to use this deleted interpolation.
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

puneet - 2018-02-09

Thanks a lot

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Log in to post a comment.