dbacl - digramic Bayesian classifier Git

104 lines (102 with data), 2.9 kB

\" t
.TH SPHERECL 1 "Bayesian Text Classification Tools" "Version @VERSION@" ""
.SH NAME
sphercl \- a spam header reweighting classifier for binary email filtering.
.SH SYNOPSIS
.HP
.B spherecl
[-@]
-C
.I category1:category2
[-l
.I category]
[FILE]...
.HP
.B spherecl
-V
.SH DESCRIPTION
.PP 
.B THIS TOOL IS NOT READY FOR USE. IT IS A WORK IN PROGRESS.
.PP
.B spherecl
is a (binary) Bayesian email classifier. 
It expects one or more FILE(s) containing single email messages in
RFC2822 format, and outputs for each a best guess category name, which
is either category1 or category2. If the 
.B -l
switch is used, spherecl learns the messages in FILE(s) as examples of
the category (which must be one of category1 or category2), rather than
classifying the messages.
.PP
spherecl in fact delegates some of its work to dbacl(1)
behind the scenes. Essentially, it modifies the dbacl scores to
retroactively give more weight to header tokens in such a way as to
try to maximize the binary classification ROC score.
.PP
Because spherecl must make certain assumptions about the dbacl output, it
contains a hardcoded list of command line options for dbacl. These options
cannot be adjusted, however it is possible to pass additional options
to dbacl if desired (not recommended).
.SH EXIT STATUS
The normal shell exit conventions aren't followed (sorry!). When using the
.B -l
command form,
.B spherecl
returns zero on success, nonzero if an error occurs. When using the
.B -c
form,
.B spherecl
returns a positive integer corresponding to the best
.IR category .
If an error occurs,
.B spherecl
returns zero.
.SH OPTIONS
.IP -@
This is a marker switch. It can be used to tell spherecl to stop analyzing
switches that follow the marker. These switches will be passed on to the 
dbacl child process verbatim. Note that spherecl expects the mandatory
.B -l
or
.B -c
switches to be used before 
.IR -@ .
Use with caution.
.IP -C
This switch is mandatory. It identifies the binary model uses by spherecl.
.IP -l
When this switch is used, spherecl considers each FILE to be a mail message
that must be learned as an example of category. Note that when this switch
is absent, spherecl will classify each FILE as one of category1 or category2.
.SH USAGE
.PP
See the documentation for dbacl regarding email filtering.
.SH ENVIRONMENT
.PP
.IP DBACL_PATH
When this variable is set, its value is prepended to every
.I category
filename which doesn't start with a '/' or a '.'.
.SH SOURCE
.PP
The source code for the latest version of this program is available at the
following locations:
.PP
.na
http://www.lbreyer.com/gpl.html
.br
http://dbacl.sourceforge.net
.ad
.SH AUTHOR
.PP
Laird A. Breyer <laird@lbreyer.com>
.SH SEE ALSO
.PP
.BR dbacl (1),
.BR bayesol (1),
.BR hmine (1),
.BR hypex (1),
.BR mailcross (1),
.BR mailfoot (1),
.BR mailinspect (1),
.BR mailtoe (1)