Menu

hub4 sphinx4

Help
Andrew
2006-07-26
2012-09-22
  • Andrew

    Andrew - 2006-07-26

    I am trying to get hub4 working with sphinx so I can eventually transcribe wavfiles about around 5 mins long and varying words spoken. I thought I would first try to get hub4 working with the latticedemo.

    I got the following output

    Loading...
    <s> <sil> <sil> what's derosa erode zero one <sil> </s>
    <s> <sil> <sil> what's derosa erode zero one </s>
    <s> <sil> <sil> what's derosa erode zero one <sil> <sil> </s>
    <s> <sil> <sil> what's derosa erode zero one <sil> </s>
    <s> <sil> <sil> what's derosa erode zero want </s>
    <s> <sil> <sil> what's derosa erodes there'll want <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll want </s>
    <s> <sil> <sil> what's derosa erodes there'll want <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll want <sil> <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll walked <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll walked </s>
    <s> <sil> <sil> what's derosa erodes there'll walked <sil> <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll walked <sil> </s>
    <s> <sil> <sil> what's derosa erodes there'll want </s>
    I heard: what's derosa erode zero one

    <s> <sil> that o'toole uno <sil> </s>
    <s> <sil> that o'toole uno </s>
    <s> <sil> that o'toole uno <sil> </s>
    <s> <sil> that o'toole uno <sil> <sil> </s>
    <s> <sil> that o'toole uno </s>
    <s> <sil> that o'toole unknowns </s>
    I heard: that o'toole uno

    <s> <sil> cyril one eight's erodes re <sil> </s>
    <s> <sil> cyril one eight's erodes re </s>
    <s> <sil> cyril one eight's erodes re <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes re <sil> </s>
    <s> <sil> cyril one eight's erodes rea </s>
    <s> <sil> cyril one eight's erodes rea <sil> </s>
    <s> <sil> cyril one eight's erodes rea <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes rea <sil> </s>
    <s> <sil> cyril one eight's erodes rhee </s>
    <s> <sil> cyril one eight's erodes rhee <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes rhee <sil> </s>
    <s> <sil> cyril one eight's erodes rhee <sil> </s>
    <s> <sil> cyril one eight's erodes ree </s>
    <s> <sil> cyril one eight's erodes ree <sil> </s>
    <s> <sil> cyril one eight's erodes ree <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes ree <sil> </s>
    <s> <sil> cyril one eight's erodes read </s>
    <s> <sil> cyril one eight's erodes read <sil> </s>
    <s> <sil> cyril one eight's erodes read <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes read <sil> </s>
    <s> <sil> cyril one eight's erodes reid </s>
    <s> <sil> cyril one eight's erodes reid <sil> </s>
    <s> <sil> cyril one eight's erodes reid <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes reid <sil> </s>
    <s> <sil> cyril one eight's erodes reed <sil> </s>
    <s> <sil> cyril one eight's erodes reed </s>
    <s> <sil> cyril one eight's erodes reed <sil> </s>
    <s> <sil> cyril one eight's erodes reed <sil> <sil> </s>
    <s> <sil> cyril one eight's erodes re </s>
    <s> <sil> cyril one eight's erodes rea </s>
    <s> <sil> cyril one eight's erodes ree </s>
    <s> <sil> cyril one eight's erodes rhee </s>
    <s> <sil> cyril one eight's erodes read </s>
    <s> <sil> cyril one eights erodes re <sil> </s>
    <s> <sil> cyril one eights erodes re </s>
    <s> <sil> cyril one eights erodes re <sil> <sil> </s>
    <s> <sil> cyril one eights erodes re <sil> </s>
    <s> <sil> cyril one eights erodes rea </s>
    <s> <sil> cyril one eights erodes rea <sil> </s>
    <s> <sil> cyril one eights erodes rea <sil> <sil> </s>
    <s> <sil> cyril one eights erodes rea <sil> </s>
    <s> <sil> cyril one eights erodes rhee </s>
    <s> <sil> cyril one eights erodes rhee <sil> <sil> </s>
    <s> <sil> cyril one eights erodes rhee <sil> </s>
    <s> <sil> cyril one eights erodes rhee <sil> </s>
    <s> <sil> cyril one eights erodes ree </s>
    <s> <sil> cyril one eights erodes ree <sil> </s>
    <s> <sil> cyril one eights erodes ree <sil> <sil> </s>
    <s> <sil> cyril one eights erodes ree <sil> </s>
    <s> <sil> cyril one eights erodes read </s>
    <s> <sil> cyril one eights erodes read <sil> </s>
    <s> <sil> cyril one eights erodes read <sil> <sil> </s>
    <s> <sil> cyril one eights erodes read <sil> </s>
    <s> <sil> cyril one eights erodes reid </s>
    <s> <sil> cyril one eights erodes reid <sil> </s>
    <s> <sil> cyril one eights erodes reid <sil> <sil> </s>
    <s> <sil> cyril one eights erodes reid <sil> </s>
    <s> <sil> cyril one eights erodes reed <sil> </s>
    <s> <sil> cyril one eights erodes reed </s>
    <s> <sil> cyril one eights erodes reed <sil> </s>
    <s> <sil> cyril one eights erodes reed <sil> <sil> </s>
    <s> <sil> cyril one eights erodes re </s>
    <s> <sil> cyril one eights erodes rea </s>
    <s> <sil> cyril one eights erodes ree </s>
    <s> <sil> cyril one eights erodes rhee </s>
    <s> <sil> cyril one eights erodes read </s>
    I heard: cyril one eight's erodes re

    <s> </s>
    I heard:

    <s> </s>
    I heard:

    the I heard silence continues as far as I know infinitly.

    Also before it executes it gives a whole bunch of warnings about missing words in the dictionary.

    Any ideas on whats going wrong?

    Thanks,
    Andrew

     
    • Darren Remington

      On the missing words, your current config.xml file is pointing to the WSJ acoustic model, not the HUB4 acoustic model. Also, I happen to know from experience that the dictionary in the HUB4 acoustic model (and the WSJ AM) is missing about 1130 words from the HUB4 LM.

      This will certainly impact the the ability to build trees based on known words.

      As for the infinite loop - not sure about that one, as I haven't run into that before.

       
    • Andrew

      Andrew - 2006-07-26

      <!-- ******** -->
      <!-- The linguist configuration -->
      <!-- ******** -->

      &lt;component name=&quot;lexTreeLinguist&quot; 
                  type=&quot;edu.cmu.sphinx.linguist.lextree.LexTreeLinguist&quot;&gt;
          &lt;property name=&quot;logMath&quot; value=&quot;logMath&quot;/&gt;
          &lt;property name=&quot;acousticModel&quot; value=&quot;wsj&quot;/&gt;
          &lt;property name=&quot;languageModel&quot; value=&quot;trigramModel&quot;/&gt;
          &lt;property name=&quot;dictionary&quot; value=&quot;dictionary&quot;/&gt;
          &lt;property name=&quot;addFillerWords&quot; value=&quot;false&quot;/&gt;
          &lt;property name=&quot;fillerInsertionProbability&quot; value=&quot;1E-10&quot;/&gt;
          &lt;property name=&quot;generateUnitStates&quot; value=&quot;false&quot;/&gt;
          &lt;property name=&quot;wantUnigramSmear&quot; value=&quot;true&quot;/&gt;
          &lt;property name=&quot;unigramSmearWeight&quot; value=&quot;1&quot;/&gt;
          &lt;property name=&quot;wordInsertionProbability&quot; 
                  value=&quot;${wordInsertionProbability}&quot;/&gt;
          &lt;property name=&quot;silenceInsertionProbability&quot; 
                  value=&quot;${silenceInsertionProbability}&quot;/&gt;
          &lt;property name=&quot;languageWeight&quot; value=&quot;${languageWeight}&quot;/&gt;
          &lt;property name=&quot;unitManager&quot; value=&quot;unitManager&quot;/&gt;
      &lt;/component&gt;
      
      
      &lt;!-- ******************************************************** --&gt;
      &lt;!-- The Dictionary configuration                            --&gt;
      &lt;!-- ******************************************************** --&gt;
      &lt;component name=&quot;dictionary&quot; 
          type=&quot;edu.cmu.sphinx.linguist.dictionary.FastDictionary&quot;&gt;
          &lt;property name=&quot;dictionaryPath&quot;
                    value=&quot;resource:/edu.cmu.sphinx.model.acoustic.WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz.Model!/edu/cmu/sphinx/model/acoustic/WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz/dict/cmudict.0.6d&quot;/&gt;
          &lt;property name=&quot;fillerPath&quot; 
                value=&quot;resource:/edu.cmu.sphinx.model.acoustic.WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz.Model!/edu/cmu/sphinx/model/acoustic/WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz/dict/fillerdict&quot;/&gt;
          &lt;property name=&quot;addSilEndingPronunciation&quot; value=&quot;false&quot;/&gt;
          &lt;property name=&quot;wordReplacement&quot; value=&quot;&amp;lt;sil&amp;gt;&quot;/&gt;
          &lt;property name=&quot;unitManager&quot; value=&quot;unitManager&quot;/&gt;
      &lt;/component&gt;
      
      
      &lt;!-- ******************************************************** --&gt;
      &lt;!-- The Language Model configuration                         --&gt;
      &lt;!-- ******************************************************** --&gt;
      &lt;component name=&quot;trigramModel&quot; 
            type=&quot;edu.cmu.sphinx.linguist.language.ngram.large.LargeTrigramModel&quot;&gt;
          &lt;property name=&quot;unigramWeight&quot; value=&quot;.5&quot;/&gt;
          &lt;property name=&quot;maxDepth&quot; value=&quot;3&quot;/&gt;
          &lt;property name=&quot;logMath&quot; value=&quot;logMath&quot;/&gt;
          &lt;property name=&quot;dictionary&quot; value=&quot;dictionary&quot;/&gt;
          &lt;property name=&quot;location&quot;
           value=&quot;C://Program Files/sphinx4/bld/models/language_model.arpaformat.DMP&quot;/&gt;
      &lt;/component&gt;
      
      
      &lt;!-- ******************************************************** --&gt;
      &lt;!-- The acoustic model configuration                         --&gt;
      &lt;!-- ******************************************************** --&gt;
      &lt;component name=&quot;wsj&quot;
                 type=&quot;edu.cmu.sphinx.model.acoustic.WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz.Model&quot;&gt;
          &lt;property name=&quot;loader&quot; value=&quot;wsjLoader&quot;/&gt;
          &lt;property name=&quot;unitManager&quot; value=&quot;unitManager&quot;/&gt;
      &lt;/component&gt;
      
      &lt;component name=&quot;wsjLoader&quot; type=&quot;edu.cmu.sphinx.model.acoustic.WSJ_8gau_13dCep_16k_40mel_130Hz_6800Hz.ModelLoader&quot;&gt;
          &lt;property name=&quot;logMath&quot; value=&quot;logMath&quot;/&gt;
          &lt;property name=&quot;unitManager&quot; value=&quot;unitManager&quot;/&gt;
      &lt;/component&gt;
      

      This is my current config.xml file

       

Log in to post a comment.