Ted Pedersen > Text-SenseClusters-1.01 > simat.pl

Download:
Text-SenseClusters-1.01.tar.gz

Annotate this POD

CPAN RT

Open  0
View/Report Bugs
Source   Latest Release: Text-SenseClusters-1.03

NAME ^

simat.pl - Build a similarity matrix from real-valued context vectors

SYNOPSIS ^

simat.pl [OPTIONS] VECTOR

The input file represents 5 vectors, each with 4 possible features. The format of the input file is sparse, so if a feature has no value it is not listed.

 cat input

Output =>

 5 4 12
 1 1.5 3 2.5 4 1.0
 2 2.5 3 2.5
 1 1.5 3 2.5 4 1.0
 2 2.5 3 2.5
 2 2.5 3 2.5

Compute the pairwise similarities between all 5 vectors and display them in a 5x5 matrix.

 simat.pl input --format f4.2

Output =>

 5 25
 1 1.00 2 0.57 3 1.00 4 0.57 5 0.57
 1 0.57 2 1.00 3 0.57 4 1.00 5 1.00
 1 1.00 2 0.57 3 1.00 4 0.57 5 0.57
 1 0.57 2 1.00 3 0.57 4 1.00 5 1.00
 1 0.57 2 1.00 3 0.57 4 1.00 5 1.00

Type simat.pl for a quick summary of options

DESCRIPTION ^

Constructs a similarity matrix for the given real-valued context vectors. A similarity matrix shows the pairwise similarities between all the different vectors. Vectors are represented in an N x M matrix, where N is the number of vectors and M is the number of features. All NxN combinations of vector pairs will be measured for similarity and stored in a matrix.

INPUT ^

Required Arguments:

VECTOR

Should be a file containing sparse/dense vectors as created by programs in dir Toolkit/vector.

Sparse Format (default)

When the input vectors are in sparse format, the first line in the VECTOR file should show 3 space separated numbers -

 N M NNZ

where

 N = Number of rows/vectors
 M = Number of columns
 NNZ = Number of non-zero values

Each line after this line should show a single sparse vector on each line. A sparse vector is a list of pairs of numbers separated by space such that the first number in a pair is the index of a non-zero value in the vector and the second number is a non-zero value itself corresponding to that index.

Column indices start with 1.

Sample Sparse Format -

 7 8 27
 3 0.727 4 0.921 7 0.734 8 0.841
 6 0.726 7 0.775 8 0.948
 1 0.898 3 0.712 5 0.787 8 0.724
 4 0.797 6 0.649
 1 0.695 2 0.710 3 0.837 4 0.736 7 0.631
 2 0.661 3 0.778 4 0.762 6 0.957
 1 0.915 3 0.639 5 0.989 6 0.637 8 0.649

Explanation :

  1. First line shows that there are 7 sparse vectors, represented using 8 features, and total 27 non-zero values.
  2. Each vector (all lines except the 1st line) is a list of 'INDEX VALUE' pairs separated by space.

    e.g. line 2 shows that the 1st vector has a non-zero value 0.727 at index 3, 0.921 at 4, value 0.734 at index 7 and 0.841 at 8.

    Only those 'INDEX VALUE' pairs are listed in which value is non-zero.

    Column indices start from 1, to be consistent with Cluto's matrix format standard.

Note that, the values could be integer and negative.

Dense Format

If VECTORs are in dense format, switch --dense should be selected.

For N vectors each having M components, the first line of dense VECTOR file should show exactly 2 integers N M i.e. number of vectors and number of components, separated in space. This should be followed by actual vectors each on a separate line. Each dense vector should only list the vector component values (and not the corresponding column indices), and should show all values including 0s. Thus, all vectors should be equal in length and should show M space separated numbers.

Sample Dense Format -

 7 8
 0.000 0.000 0.727 0.921 0.000 0.000 0.734 0.841
 0.000 0.000 0.000 0.000 0.000 0.726 0.775 0.948
 0.898 0.000 0.712 0.000 0.787 0.000 0.000 0.724
 0.000 0.000 0.000 0.797 0.000 0.649 0.000 0.000
 0.695 0.710 0.837 0.736 0.000 0.000 0.631 0.000
 0.000 0.661 0.778 0.762 0.000 0.957 0.000 0.000
 0.915 0.000 0.639 0.000 0.989 0.637 0.000 0.649

Shows same VECTOR file as shown in section Sample Sparse Format, in dense format.

VECTOR file could also optionally show the <keyfile> tag on the first line. If the first line shows the <keyfile> name, then the above VECTOR description should start from the 2nd line onwards.

Optional Arguments:

--dense

This switch should be selected when the given VECTORs are in dense format. This will also create the output similarity matrix in dense format. By default, sparse format is assumed and used for both input vectors and output similarity matrix.

--format FORM

Specifies the numeric representation format for output similarity matrix.

Acceptable FORM value is

fN.M -> each similarity value is represented as a floating point number occupying total N byte space with last M bytes showing the fractional part.

Default format used is f16.10.

Since the cosine similarity values computed by simat are always in the range [0-1], simat supports only floating point format.

Other Options :

--help

Displays this message.

--version

Displays the version information.

OUTPUT ^

Sparse Format [default]

By default (when --dense is not selected), OUTPUT will be created in sparse format. In sparse format, the 1st line will show two space separated numbers,

 N NNZ1 

where N = Number of vectors, same as the N in the given VECTOR file NNZ1 = Number of non-zero values in the output similarity matrix

Each i'th line after the above line shows the list of 'j COSINE' pairs separated by space such that COSINE is the non-zero cosine similarity value between the i'th and j'th vectors in the given VECTOR file.

Only those pairs are listed in which the COSINE values are non-zeroes.

Sample Sparse Output

 7 29
 1 1.000 4 0.441 6 0.491 7 0.357
 2 1.000 3 0.308 4 0.322 5 0.212
 2 0.308 3 1.000
 1 0.441 2 0.322 4 1.000 5 0.352 6 0.750 7 0.230
 2 0.212 4 0.352 5 1.000 6 0.651
 1 0.491 4 0.750 5 0.651 6 1.000 7 0.398
 1 0.357 4 0.230 6 0.398 7 1.000

Shows the actual full similarity matrix -

 7
 1.000 0.000 0.000 0.441 0.000 0.491 0.357
 0.000 1.000 0.308 0.322 0.212 0.000 0.000
 0.000 0.308 1.000 0.000 0.000 0.000 0.000
 0.441 0.322 0.000 1.000 0.352 0.750 0.230
 0.000 0.212 0.000 0.352 1.000 0.651 0.000
 0.491 0.000 0.000 0.750 0.651 1.000 0.398
 0.357 0.000 0.000 0.230 0.000 0.398 1.000

Both the matrices show the pair-wise similarities among 7 vectors. e.g. Row 1 shows that the cosine similarity among the 1st vector and the 4th vector is 0.441, between 1st and 6th is 0.491, and the similarity between 1st and 7th vectors is 0.357. In sparse format, each line shows only those values that are non-zeroes.

Note that, the similarity matrix (sparse/dense) always represents a square symmetric matrix.

Dense Format

For N input vectors, the first line of the dense output shows a single integer number N (the total number of vectors). Thereafter, each i'th line shows pair-wise similarities of i'th vector with each of the N vectors. i.e. each row contains N columns such that the value at j'th column in i'th row shows the pair-wise similarity between i'th and j'th vectors.

Sample Dense Output

 7
 1.000 0.000 0.000 0.441 0.000 0.491 0.357
 0.000 1.000 0.308 0.322 0.212 0.000 0.000
 0.000 0.308 1.000 0.000 0.000 0.000 0.000
 0.441 0.322 0.000 1.000 0.352 0.750 0.230
 0.000 0.212 0.000 0.352 1.000 0.651 0.000
 0.491 0.000 0.000 0.750 0.651 1.000 0.398
 0.357 0.000 0.000 0.230 0.000 0.398 1.000

where each cell shows the similarity between the corresponding pair of vectors.

SYSTEM REQUIREMENTS ^

simat.pl is dependent on the following CPAN modules :

PDL - http://search.cpan.org/dist/PDL/
Math::SparseVector - http://search.cpan.org/dist/Math-SparseVector/

AUTHORS ^

 Amruta Purandare, University of Pittsburgh

 Ted Pedersen, University of Minnesota, Duluth
 tpederse at d.umn.edu

COPYRIGHT ^

Copyright (c) 2002-2008, Amruta Purandare and Ted Pedersen

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program; if not, write to

 The Free Software Foundation, Inc.,
 59 Temple Place - Suite 330,
 Boston, MA  02111-1307, USA.
syntax highlighting: