svdinterface.1

有关自然语言理解理解方面的源码

.\" Process this file with 
.\"    groff -man -Tascii svdinterface.1

.TH SVDINTERFACE 1 "February 2004" "Infomap Project" "Infomap NLP Manual"

.SH NAME
.TP 
svdinterface \- invoke SVDPACKC to perform SVD

.SH SYNOPSIS
.B svdinterface
.BR -singvals " <num_singvals> " -iter " <num_iter> "

.SH DESCRIPTION
.B svdinterface
uses the SVDPACKC library from the University of Tennessee to reduce
the dimensions of a matrix using Singular Value Decomposition (SVD).
Note that the SVDPACKC library (and therefore the 
.B svdinterface
program) are under a different license from the rest of the 
Infomap NLP software.
In particular, the incorporation of SVDPACKC or the
.B svdinterface 
program in a commercial product requires the express consent of
the University of Tennessee.  You have been warned.

Note that 
.B svdinterface
must be run in the directory in which the model is being built.
(I.e. the 
.IR "model data directory" ,
which contains the actual model files.)  The
.B infomap-build 
program takes care of this requirement automatically.

.SH OPTIONS
.TP
.BI -singvals \ <num_singvals>
The number of singular values to be produced in the resulting
matrix.  This is therefore the number of dimensions that the
word vectors will ultimately have.

.TP
.BI -iter \ <num_iter>
The number of iterations to use in the algorithm that computes the
SVD.

.\" .SH EXAMPLES

.SH INPUT FILES
These files are read from the model data directory, which must
be the current working directory when 
.B svdinterface
is invoked.

.I indx
.RS
This file and 
.I coll
taken together encode the pre-SVD matrix.  See
.BR count_wordvec (1).
.RE

.I coll
.RS 
This file and
.I indx
taken together encode the pre-SVD matrix.  See
.BR count_wordvec (1).
.RE

.I model_params.bin
.RS
Reads in this file and writes out a modified version.
See 
.BR prepare_corpus (1).
.RE

.I model_info.bin
.RS
Reads in this file and writes out a modified version.
See
.BR prepare_corpus (1).
.RE


.SH OUTPUT FILES
All of these files are written to the model data directory,
which is taken to be the current working directory when
.B svdinterface
is invoked.

.I left
.RS
The left singular vectors that result from the Singular Value
Decomposition.  These are the "word vectors" that are used by
.BR associate (1).
.RE

.I rght
.RS
The right singular vectors that result from the Singular Value
Decomposition.  These are not used by the current implementation.
.RE

.I sing
.RS
The singular values produced by Singular Value Decomposition (the
diagonal elements of the diagonal matrix that is produced).  These
are not used by the current implementation.
.RE

.I model_params.bin
.RS
Reads in this file and writes out a modified version.
See 
.BR prepare_corpus (1).
.RE

.I model_info.bin
.RS
Reads in this file and writes out a modified version.
See
.BR prepare_corpus (1).
.RE


.SH SEE ALSO
.BR prepare_corpus (1), \ count_wordvec (1), \ encode_wordvec (1), \
\ count_artvec (1), \ write_text_params (1).

.SH DIAGNOSTICS
Returns 0 to indicate success; nonzero value to indicate error.

.SH BUGS
This page should have a more detailed description of the input
and output file formats.  Ideally, we should have a single interface
program that can use various different SVD libraries on the back end.
Please report bugs to 
.BR [e-mail] .

.SH CREDITS
The Infomap NLP software was written by Stefan Kaufmann, Hinrich
Schuetze, Dominic Widdows, Beate Dorow, and Scott Cederberg.  The
Infomap algorithm was originally developed by Hinrich Schuetze.  
.B svdinterface
uses the SVDPACKC library from the University of Tennessee.
Incorporating this library in a commercial product requires the 
express permission of the University of Tennessee.

.SH AUTHOR
This manual page was written by Scott Cederberg.  Please direct
inquiries and bug reports to 
.BR [e-mail] .