selectd.8

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。

.\"Generated by db2man.xsl. Don't modify this, modify the source.
.de Sh \" Subsection
.br
.if t .Sp
.ne 5
.PP
\fB\\$1\fR
.PP
..
.de Sp \" Vertical space (when we can't use .PP)
.if t .sp .5v
.if n .sp
..
.de Ip \" List item
.br
.ie \\n(.$>=3 .ne \\$3
.el .ne 3
.IP "\\$1" \\$2
..
.TH "SELECTD" 8 "" "" ""
.SH NAME
selectd \- email classification daemon
.SH "SYNOPSIS"

.nf
\fBselectd\fR [\fB-a \fIaddress\fR\fR] [\fB-f \fIconfigfile\fR\fR]
.fi

.SH "DESCRIPTION"

.PP
To be written\&.

.SH "COMMAND LINE OPTIONS"

.TP
\fB-a\fR or \fB--address\fR

Selects socket address to use\&.
This overrides the value given in the configuration file\&.


.TP
\fB-f\fR or \fB--conf-file\fR

Selects which configuration file to use\&.
Default value is \fI$HOME/\&.select/selectd\&.conf\fR\&.


.SH "CONFIGURATION"

.PP
Configuration options for selectd is given in a file\&. The default name of this file is \fI$HOME/\&.select/selectd\&.conf\fR\&.

.SS "Location options"

.PP
Options describing location of directories etc\&.

.TP
"address"

.PP
Specifies the daemon address\&. Addresses beginning with the character @ are interpreted as abstract addresses, otherwise they are filename addresses\&. Default value is @my_select_classifier\&. 

.nf
address                                 # Default address
address qwerty                          # Use the specified address

.fi
 

.TP
"data_dir"

.PP
Data directory\&. Indicates which directory database files are stored in\&. Default value is to not save databases\&. 

.nf
data_dir                                # Default: Don't save databases
data_dir        /home/mikael/\&.select    # Use the specified directory

.fi
 

.TP
"share_dir"

.PP
Share directory\&. Indicates which directory language files are stored in\&. Default value is \fI/usr/local/share/select\fR\&. 

.nf
share_dir                               # Default: /usr/local/share/select
share_dir       share                   # Use the specified directory

.fi
 

.TP
"plugin_dir"

.PP
Plugin directory\&. Indicates which directory classifiers and stemmers are stored in\&. Default value is \fI/usr/local/lib/select\fR\&. 

.nf
plugin_dir                              # Default: /usr/local/lib/select
plugin_dir      plugins                 # Use the specified directory

.fi
 

.SS "Language options"

.PP
Options describing language handling\&.

.TP
"language"

.PP
Specifies a language\&. May be present several times\&. The first presence sets the default language\&. Default value is no languages\&.

.PP
A language is specified by its name followed by a default locale string\&. The name must be the same as is used by its stopwords and stemmer\&. The stopwords and stemmers distibuted with Select use the language codes specified in ISO 639 as names\&. If you use a custom stopword list (for example for several languages) a new name should be invented for it\&. 

.nf
language                                # No languages
language        en C                    # English
language        sv sv_SE                # Swedish
language        sven sv_SE              # Swenglish

.fi
 

.TP
"identifier"

.PP
Language identification\&. Default value is off\&. 

.nf
identifier      on                      # Use language identification
identifier      off                     # Don't use language identification

.fi
 

.TP
"stemming"

.PP
Stemming\&. Default value is off\&. 

.nf
stemming        on                      # Perform stemming for chosen languages
stemming        off                     # Don't perform stemming

.fi
 

.TP
"stopwords"

.PP
Stopwords\&. Default value is off\&. 

.nf
stopwords       on                      # Use stopwords for chosen languages
stopwords       off                     # Don't use any stopwords

.fi
 

.SS "Classifier Options"

.PP
Classifiers are configured in special classifier sections in the configuration file\&. A classifier selection is started by a line like this: 

.IP
.nf
[classifier]

.fi
 There may be several classifiers specified\&.

.TP
"name"

.PP
Classifier name\&. No default value\&. 

.nf
name            Henry                   # Call the classifier Henry

.fi
 

.TP
"classifier"

.PP
Classifier algorithm\&. No default value\&. 

.nf
# Multi classifiers
classifier      Alma                    # Alma classifier
classifier      NaiveBayes              # Naive bayes classifier
classifier      Perceptron              # Perceptron classifier
classifier      Rocchio                 # Rocchio classifier
classifier      Romma                   # Romma classifier
classifier      WidrowHoff              # Widrow-Hoff (LMS) classifier
classifier      WMA                     # Weighted Majority Algorithm
# Document classifiers
classifier      From                    # Sender classifier
classifier      Reply                   # Thread classifier

.fi
 

.TP
"type"

.PP
Classifier type\&. No default value\&. 

.nf
type            multi_one               # Multi classifier, type ONE_MAX
type            multi_rest              # Multi classifier, type REST_MAX
type            multi_linmax            # Multi classifier, type LIN_MAX
type            multi_uc                # Multi classifier, type UC_MAX
type            document                # Document classifier

.fi
 

.TP
"options"

.PP
Classifier options\&. A string which is passed on to the classifier for interpretation\&. Is used for changing the behavior of the classifier\&. The string consists of a comma-separated list of entries of the form key=value\&. Valid keys and values depends on the classifier\&. Default value is none\&. 

.nf
options                                 # No options
options         n=200,o=fifo            # Classifier dependent options

.fi
 

.TP
"tokenizer"

.PP
Tokenizer\&. Specifies which method is used for transforming texts into text-tokens\&. Default value is alpha\&. 

.nf
tokenizer       alpha                   # Alpha tokenizer
tokenizer       wspace\&.byte             # Whitespace byte tokenizer
tokenizer       ngram\&.byte              # N-gram byte tokenizer
tokenizer       null                    # Null tokenizer

.fi
 

.TP
"vectorizer"

.PP
Vectorization mode\&. Specifies which method is used for transforming text-tokens into a vector\&. Default value is tf\&. 

.nf
vectorizer      bool                    # Boolean
vectorizer      tf                      # Term Frequency
vectorizer      tfidf                   # TF-IDF

.fi
 

.TP
"normalizer"

.PP
Normalization mode\&. Specifies how vectors should be normalized\&. Default value is none\&. 

.nf
normalizer                              # No normalization
normalizer      cosine                  # Cosine normalization

.fi
 

.SS "Example"

.PP
Here is a complete example of a selectd configuration file:

.IP
.nf

.nf
address                                 # Use default address
data_dir        /home/mikael/\&.select    # Save databases in specified directory
share_dir                               # Default: /usr/local/share/select
plugin_dir                              # Default: /usr/local/lib/select

language        sven sv_SE              # Swenglish
identifier      off                     # Don't use language identification
stemming        off                     # Don't perform stemming
stopwords       on                      # Use stopwords for chosen languages

[classifier]
name            nb                      # Name the classifier nb
classifier      NaiveBayes              # NaiveBayes classifier
type            multi_one               # Multi classifier, type ONE_MAX
options                                 # No options
tokenizer       alpha                   # Alpha tokenizer
vectorizer      tfidf                   # TF-IDF vectorizer
normalizer                              # No normalization

.fi

.fi

.SH "AUTHOR"

.PP
Mikael Ylikoski <my@post\&.utfors\&.se>