Menu

Using a new model in pocketsphinx

Help
Tamirru
2013-12-23
2017-05-18
  • Tamirru

    Tamirru - 2013-12-23

    Hi,
    I am new at CMU Sphinx.
    My current argument setup of pocketsphinx is with as follow:
    -hmm ../../model/hmm/wsj
    -lm ../../model/lm/en/language_model.arpaformat.DMP
    -dict ../../model/lm/en/dic.dic

    I use windows installation of pocketsphinx,
    and I want to use this model:
    http://cmusphinx.sourceforge.net/2013/01/a-new-english-language-model-release/
    where I've downloaded 2 files: en-70k.lm and en-70k-0.1.lm.gz.md5

    what should I do to use this model with my pocketsphinx setup?
    Also, do I need to use specific audio input parameters? (frequency, mono etc.)

    Thx!

     

    Last edit: Tamirru 2013-12-23
  • Tamirru

    Tamirru - 2013-12-23

    Thx, with the dmp file it is working, But I would like to use en-70k.lm
    how should I run it? I tried -lm (instead of the dmp file) but got blank output

     
  • Nickolay V. Shmyrev

    You can not use 70k lm with pocketsphinx, you need to limit vocabulary to less than 60k (or to 20k like in dmp file).

     
    • kunal sharma

      kunal sharma - 2017-05-18

      how can i limit the vocabulary in lm file ?

       
      • Nickolay V. Shmyrev

        This thread is not relevant anymore, pocketsphinx should work with the model above.

         
        • misha mihail

          misha mihail - 2017-05-18

          Thank you for your reply, Nickolay. I only want few words like (start,
          stop, next, back on my video player), So can i use
          android-procketsphinx-demo with new.lm and new.dic? Or should i have to
          build dmp file?
          Please leave a message for me. Thank you!

           

          Last edit: Nickolay V. Shmyrev 2017-05-18
  • Tamirru

    Tamirru - 2013-12-23

    Ohh.. ok thanks Nickolay!
    Can you recommend on another good model for video transcript? (don't care about real time performance, I need high success rate)

     
  • Nickolay V. Shmyrev

    I've just recommended you the good models in the first reply.

     
  • Tamirru

    Tamirru - 2013-12-23

    Thought maybe video transcript might have a specific model.
    OK I'll try it out. Thx again!

     
  • rupy

    rupy - 2014-05-21

    So I tried this data, and it works better... but still only one or two words at the time and very slow. Is there anything that could improve speed and length of sentences?

     

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.