ref_select_test.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。

<refmeta>
<refentrytitle>select_test</refentrytitle>
<manvolnum>1</manvolnum>
</refmeta>

<refnamediv>
<refname>select_test</refname>
<refpurpose>email classification testing tool</refpurpose>
</refnamediv>

<refsynopsisdiv>
<cmdsynopsis>
<command>select_test</command>
<arg><option>-a <replaceable>address</replaceable></option></arg>
<arg><option>-b <replaceable>mailboxfile</replaceable></option></arg>
<arg><option>-c <replaceable>classeq</replaceable></option></arg>
<arg><option>-f <replaceable>configfile</replaceable></option></arg>
<arg><option>-m <replaceable>msgseq</replaceable></option></arg>
<arg><option>-p <replaceable>plottype</replaceable></option></arg>
<arg><option>-r <replaceable>ranklimit</replaceable></option></arg>
<arg><option>-t <replaceable>train-mode</replaceable></option></arg>
<arg><option>-C</option></arg>
<arg><option>-P <replaceable>clsnum</replaceable></option></arg>
</cmdsynopsis>
</refsynopsisdiv>

<refsect1>
<title>Description</title>
<para>
To be written.
</para>
</refsect1>

<refsect1>
<title>Command Line Options</title>

<variablelist>

<varlistentry>
<term><option>-a</option> or <option>--address</option></term>
<listitem>
<simpara>
Selects socket address to use.
This overrides the value given in the configuration file.
</simpara>
</listitem>
</varlistentry>

<varlistentry>
<term><option>-b</option> or <option>--mailbox-file</option></term>
<listitem>
<simpara>
Selects which mailbox configuration file to use.
</simpara>
</listitem>
</varlistentry>

<varlistentry>
<term><option>-f</option> or <option>--conf-file</option></term>
<listitem>
<simpara>
Selects which configuration file to use.
Default value is <filename>$HOME/.select/selectd.conf</filename>.
</simpara>
</listitem>
</varlistentry>

</variablelist>

</refsect1>

<refsect1>
<title>Configuration</title>

<para>
The behaviour of <application>select_test</application> can be changed in various ways.
This is specified in a configuration file.
</para>

<refsect2>
<title>Location options</title>
<para>
Options describing location of directories etc.
</para>

<refsect3>
<title>address</title>
<para>
Specifies the daemon address.
Addresses beginning with '<userinput>@</userinput>' are interpreted as abstract addresses, otherwise they are filename addresses.
Default value is @my_select_classifier.
<programlisting>address                                 # Default address
address qwerty                          # Use the specified address
</programlisting>
</para>
</refsect3>

<refsect3>
<title>mailbox_file</title>
<para>
Mailbox file.
This is a file specifying the mailboxes to use.
Each line of the file contains the name of a directory (containing messages) or a mbox-file.
Default value is select_test.mail.
<programlisting>mailbox_file    data/select.mail        # Use the specified file
</programlisting>
</para>
</refsect3>

<refsect3>
<title>combiner</title>
<para>
Combiner classifier.
Default value is off.
<programlisting>combiner        on                      # Use combiner
</programlisting>
</para>
</refsect3>

<refsect3>
<title>rank_limit</title>
<para>
Rank limit.
Default value is 3.
<programlisting>rank_limit      3                       # Use 3 as rank limit
</programlisting>
</para>
</refsect3>

</refsect2>

<refsect2>
<title>Training options</title>
<para>
</para>

<refsect3>
<title>class_seq</title>
<para>
Class sequence mode.
Default value is cross.
<programlisting>class_seq       cross                   #
class_seq       linear                  # Linear class sequence
class_seq       random                  # Random class sequence
class_seq       time                    # Chronological class sequence
</programlisting>
</para>
</refsect3>

<refsect3>
<title>msg_seq</title>
<para>
Message sequence mode.
Default value is linear.
<programlisting>msg_seq         linear                  # Linear message sequence
msg_seq         random                  # Random message sequence
</programlisting>
</para>
</refsect3>

<refsect3>
<title>training_mode</title>
<para>
Training mode.
Default value is interleaved.
<programlisting>training_mode   all                     #
training_mode   interleaved             #
training_mode   class_percentage        # 
training_mode   total_percentage        #
</programlisting>
</para>
</refsect3>

</refsect2>

<refsect2>
<title>Plotting options</title>
<para>
</para>

<refsect3>
<title>plot</title>
<para>Plotting mode.
Default value is last_n.
<programlisting>plot            last_n                  # Plot over the last n messages
plot            total                   # Plot over all previous messages
plot            off                     # No plot
</programlisting>
</para>
</refsect3>

<refsect3>
<title>plot_classifier</title>
<para>Plotted classifier.
Default value is 0.
<programlisting>plot_classifier 0                       # Plot classifier 0
</programlisting>
</para>
</refsect3>

</refsect2>

<refsect2>
<title>Example</title>
<para>Here is a complete example of a <application>select_test</application> configuration file:</para>
<informalexample>
<programlisting>mailbox_file    select.mail             # Read mailboxes from select.mail
rank_limit      3                       # Use 3 as rank limit

class_seq       time                    # Chronological class sequence
msg_seq         linear                  # Linear message sequence
training_mode   interleaved             #

plot            last_n                  # Plot over the last n messages
plot_classifier 0                       # Plot classifier 0
</programlisting>
</informalexample>
</refsect2>

</refsect1>