Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Word Level LM at Line 96 #40

Open
kmohanku opened this issue Nov 9, 2018 · 0 comments
Open

Word Level LM at Line 96 #40

kmohanku opened this issue Nov 9, 2018 · 0 comments

Comments

@kmohanku
Copy link

kmohanku commented Nov 9, 2018

I have a KenLM scoring integrated at Line 96. The performance on my test set (Both LM and Test set are LibriSpeech based) is worse than not using an LM at all. I am scoring only at space, multiplying the log probability (Converted from log10) by Alpha and also compensating with bonus term by adding (beta * log(word count in prefix)). I am applying this only to "not blank" probability. I have no success. Has anyone achieved success by integrating LM scoring?

I used my test set and Language model with Paddle Paddle decoder with same acoustic model and there was a 6% improvement in WER. They have a trie based LM aided by WFST correction along with this beam search algo. I would appreciate any pointers or help here. Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant