Menu

Why are the SER and WER so high?

Help
Charles Li
2007-10-25
2012-09-22
  • Charles Li

    Charles Li - 2007-10-25

    I'm using SPHINX 3 and TIMIT corpus.
    I trained the model using TIMIT train data and then decoded the TIMIT test data.
    The SER is 80% and the WER is 60%.
    Why are the SER and WER so high? Is it normal or is there anything wrong?
    Thanks.

     
    • Charles Li

      Charles Li - 2007-10-29

      Hi, Nickolay, Thanks very much for your comments. It really helps. Charles.

       
    • Nickolay V. Shmyrev

      It's fine if you are using standard language model. Timit is very bad for language modelling. See recent thread on it:

      https://sourceforge.net/forum/forum.php?thread_id=1827167&forum_id=382337

       

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.