fileformats.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="fileformats">
<title>File Formats</title>

All files produced by <application>Select</application> should be in UTF-8 encoding.

<sect1>
<title>Dictionary</title>
<para>
Dictionaries look like this when saved:
<screen>9237;264=5:hello;7562=4:luck;4=7:umbrella; ... 92=8:bookcase;
</screen>
The first number indicates the number of items in the dictionary.
</para>
</sect1>

<sect1>
<title>Vector</title>
<para>
The vector (12, 0.34, 0, 0, 34.1200, 6, 6, 6, 6, 0.12) is represented as:
<screen>8;0=12,0.34;4=34.12,6*4,0.12;
</screen>
The first number indicates the number of nonzero items in the vector.
</para>
</sect1>

<sect1>
<title>Vectorizer</title>

<para>
A boolean vectorizer:
<screen>
vectorizer bool
tokenizer alpha
normalizer
autobias 1
dictionary <replaceable>dictionary</replaceable>
</screen>
</para>

<para>
A term frequency vectorizer:
<screen>
vectorizer tf
tokenizer alpha
normalizer
autobias 1
dictionary <replaceable>dictionary</replaceable>
</screen>
</para>

<para>
A TF-IDF vectorizer:
<screen>
vectorizer tfidf
tokenizer alpha
normalizer
autobias 1
dictionary <replaceable>dictionary</replaceable>
df <replaceable>vector</replaceable>
</screen>

</para>
</sect1>

<sect1>
<title>Classifier</title>
<para>
A naive bayes classifier:
<screen>
classifier naivebayes
type multi_one
noc 4			# Number of classes
[global]		# Global section: read by load_db
nod 578
now 9912
[0]			# Class section: read by load_class
nod 154
now 3359
tf <replaceable>vector</replaceable>
[1]
nod 259
now 6527
tf <replaceable>vector</replaceable>
[2]
nod 53
now 923
tf <replaceable>vector</replaceable>
[3]
nod 112
now 2534
tf <replaceable>vector</replaceable>
</screen>
</para>
</sect1>

<sect1>
<title>Folder database</title>
<para>
<screen>
folders <replaceable>dictionary</replaceable>
</screen>
</para>

</sect1>

</chapter>