<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">On 12/3/2013 12:50 PM, Laatar Rim
      wrote:<br>
    </div>
    <blockquote
      cite="mid:1386103837.59641.YahooMailNeo@web173205.mail.ir2.yahoo.com"
      type="cite">
      <div style="color:#000; background-color:#fff; font-family:times
        new roman, new york, times, serif;font-size:12pt">in the class
        format: <br>
        <pre><i>class</i> [<i>p</i>] <i>word1</i> <i>word2</i> ...</pre>
        how can i calculate p ?<br>
      </div>
    </blockquote>
    <br>
    Use replace-words-with-classes with the outfile=  option.   This is
    explained <a
      href="http://www.speech.sri.com/pipermail/srilm-user/2007q2/000445.html">in
      a previous post</a>.  <br>
    <br>
    Andreas<br>
    <br>
    <blockquote
      cite="mid:1386103837.59641.YahooMailNeo@web173205.mail.ir2.yahoo.com"
      type="cite">
      <div style="color:#000; background-color:#fff; font-family:times
        new roman, new york, times, serif;font-size:12pt"><br>
        <div style="display: block;" class="yahoo_quoted">
          <div style="font-family: times new roman, new york, times,
            serif; font-size: 12pt;">
            <div style="font-family: HelveticaNeue, Helvetica Neue,
              Helvetica, Arial, Lucida Grande, sans-serif; font-size:
              12pt;">
              <div dir="ltr"> <font face="Arial" size="2"> Le Mardi 3
                  décembre 2013 18h18, Andreas Stolcke
                  <a class="moz-txt-link-rfc2396E" href="mailto:stolcke@icsi.berkeley.edu"><stolcke@icsi.berkeley.edu></a> a écrit :<br>
                </font> </div>
              <div class="y_msg_container">
                <div id="yiv2709919316">
                  <div>
                    <div class="yiv2709919316moz-cite-prefix">On
                      12/3/2013 1:35 AM, Laatar Rim wrote:<br
                        clear="none">
                    </div>
                    <blockquote type="cite">
                      <div
                        style="color:#000;background-color:#fff;font-family:times
                        new roman, new york, times,
                        serif;font-size:12pt;">
                        <div><span>hello, </span></div>
                        <div style="color:rgb(0, 0,
                          0);font-size:16px;font-family:'times new
                          roman', 'new york', times,
                          serif;background-color:transparent;font-style:normal;"><span><br
                              clear="none">
                          </span></div>
                        <div style="background-color:transparent;"><span>on
                            the internet I found this: <br clear="none">
                          </span></div>
                        <div style="font-family:'Times New
                          Roman';font-size:medium;"> to build and use a
                          simple class language model:</div>
                        <div style="font-family:'Times New
                          Roman';font-size:medium;">Induce classes:<br
                            clear="none">
                        </div>
                        <div style="background-color:transparent;"><span></span></div>
                        <pre>ngram-class -vocab vocab_file \
            -text input_file \
            -numclasses num \
            -class-counts output.class-counts \
            -classes output.classes </pre>
                        <pre>in this exemple we need only number of class, how can i use corpus of class ??? </pre>
                      </div>
                    </blockquote>
                    The steps for building a class-based LM are:<br
                      clear="none">
                    <br clear="none">
                    1. prepare class definition file in the format
                    described in the <br clear="none">
                       classes-format(5) manual page.  this can be done
                    by hand or from other<br clear="none">
                       knowledge sources, or automatically using word
                    clustering algorithms<br clear="none">
                       (see ngram-class(1)).<br clear="none">
                    <br clear="none">
                    2. condition the training data or counts to replace
                    words with class labels,<br clear="none">
                       using the "replace-words-with-classes" filter
                    (see training-scripts(1) <br clear="none">
                       man page).<br clear="none">
                    <br clear="none">
                    3. run ngram-count on the result of step 2.
                    <div class="yiv2709919316yqt9039324517"
                      id="yiv2709919316yqtfd85124"><br clear="none">
                      <br clear="none">
                      Andreas<br clear="none">
                      <br clear="none">
                      <br clear="none">
                    </div>
                  </div>
                </div>
                <br>
                <br>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>