<sect1 id='scfg-make-manual'>
	<title><command>scfg_train</command> <emphasis>Train the parameters of a stochastic context free grammar</emphasis></title>

    <toc depth='1'></toc>
    <para>
    </para>
    <sect2>
      <title>Synopsis</title>
      <para>
      </para>
        <!-- /amd/projects/festival/versions/v_mpiro/speech_tools_linux/bin/scfg_train -sgml_synopsis -->
        <para>
<cmdsynopsis><command>scfg_train</command>[options<arg>-grammar <replaceable>ifile</replaceable></arg>
<arg>-corpus <replaceable>ifile</replaceable></arg>
<arg>-method <replaceable>string</replaceable> " {inout}"</arg>
<arg>-passes <replaceable>int</replaceable> " {50}"</arg>
<arg>-startpass <replaceable>int</replaceable> " {0}"</arg>
<arg>-spread <replaceable>int</replaceable></arg>
<arg>-checkpoint <replaceable>int</replaceable></arg>
<arg>-heap <replaceable>int</replaceable> " {210000}"</arg>
<arg>-o <replaceable>ofile</replaceable></arg>
</cmdsynopsis>
        </para>
        <!-- DONE /amd/projects/festival/versions/v_mpiro/speech_tools_linux/bin/scfg_train -sgml_synopsis -->
      <para>

scfg_train takes a stochastic context free grammar (SCFG) and trains
the probabilities with repsect to a given bracket corpus using the
inside-outside algorithm.   This is basically an implementation
of Pereira and Schabes 1992.
Note using this program properly may require months of CPU time.
      </para>
    </sect2>
    <sect2>
      <title>OPTIONS</title>
      <para>
      </para>
        <!-- /amd/projects/festival/versions/v_mpiro/speech_tools_linux/bin/scfg_train -sgml_options -->
        <para>
<variablelist>
<varlistentry><term>-grammar</term>
<LISTITEM><PARA>
<replaceable>ifile</replaceable>

Grammar file, one rule per line. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-corpus</term>
<LISTITEM><PARA>
<replaceable>ifile</replaceable>

Corpus file, one bracketed sentence per line. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-method</term>
<LISTITEM><PARA>
<replaceable>string</replaceable>
 " {inout}"
Method for training: inout. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-passes</term>
<LISTITEM><PARA>
<replaceable>int</replaceable>
 " {50}"
Number of training passes. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-startpass</term>
<LISTITEM><PARA>
<replaceable>int</replaceable>
 " {0}"
Starting at pass N. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-spread</term>
<LISTITEM><PARA>
<replaceable>int</replaceable>

Spread training data over N passes. 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-checkpoint</term>
<LISTITEM><PARA>
<replaceable>int</replaceable>

Save grammar every N passes 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-heap</term>
<LISTITEM><PARA>
<replaceable>int</replaceable>
 " {210000}"
Set size of Lisp heap, needed for large corpora 
</PARA></LISTITEM>
</varlistentry>

<varlistentry><term>-o</term>
<LISTITEM><PARA>
<replaceable>ofile</replaceable>

Output file for trained grammar. </PARA></LISTITEM>
</varlistentry>
</variablelist>
        </para>
        <!-- DONE /amd/projects/festival/versions/v_mpiro/speech_tools_linux/bin/scfg_train -sgml_options -->
    </sect2>
  </sect1>