Difference between revisions of "2013-10-18"

From cslt Wiki
Jump to: navigation, search
(以内容“== Data sharing == * LM count files still undelivered! == DNN progress == === Sparse DNN === * Optimal Brain Damage(OBD). Code ready, bug found. === Tencent exps ...”创建新页面)
 
(Noisy training)
Line 16: Line 16:
  
 
1. With 863 clean test, by adding car & white noise at various levels, obtained significant performance improvement.
 
1. With 863 clean test, by adding car & white noise at various levels, obtained significant performance improvement.
2. The test with both car & white noise benefits from the noisy training.
 
  
 +
2. The test with both car & white noise benefits from the noisy training.
  
 
==Continuous LM ==
 
==Continuous LM ==

Revision as of 01:26, 18 October 2013

Data sharing

  • LM count files still undelivered!

DNN progress

Sparse DNN

  • Optimal Brain Damage(OBD). Code ready, bug found.

Tencent exps

N/A


Noisy training

1. With 863 clean test, by adding car & white noise at various levels, obtained significant performance improvement.

2. The test with both car & white noise benefits from the noisy training.

Continuous LM

1. Lattice rescoring toolkit is ready. 2. Rescoring is slow with some dense lattices.


QA LM

1. use the QA word segment system 2. train the Q LM & QA ASR system