utilities.xml

使用具有增量学习的监控式学习方法。包括几个不同的分类算法。

<chapter id="utilities">
<title>Utility Programs</title>

<para>
The <application>Select</application> distribution includes some utility programs.
</para>


<sect1>
<title>select_analysis</title>

<para>
Prints all vectors in a classifier database with terms sorted in order of weight.
This makes it easy to see which words are most important for each class, and could maybe be used for finding good stopwords.
</para>

<sect2>
<title>Usage</title>
<screen>select_analysis &lt;vectorizer file&gt; &lt;database file&gt;
</screen>
<para>

</para>

</sect2>

</sect1>


<sect1>
<title>select_mailstat</title>

<para>
Shows some statistics about mailboxes.
</para>

</sect1>


<sect1 id="select_textstat">
<title>select_textstat</title>

<para>
Counts n-grams from a text.
These can be used for language identification.
</para>

</sect1>


<sect1 id="pifiled">
<title>pifiled</title>

<para>
<application>pifiled</application> is a daemon which imitates the behaviour of <application>selectd</application>, but runs <application>ifile</application> as a back-end
<footnote><para>It is even possible to run <application>select_ifile</application> as a back-end if you make <application>select_ifile</application> and <application>selectd</application> use another socket address than the one <application>pifiled</application> uses</para></footnote>.
</para>

<para>
This makes it possible to run <application>Select</application> clients against the <application>ifile</application> database. This is useful for testing purposes, comparing the performance of <application>ifile</application> and <application>Select</application> using <application>select_test</application>. Using this for other purposes is discouraged, as it involves unnecessary overhead.
</para>

</sect1>


</chapter>