Text::NSP::Measures::2D::MI.3pm

Langue: en

Version: 2006-06-21 (mandriva - 01/05/08)

Section: 3 (Bibliothèques de fonctions)

Sommaire

NAME

Text::NSP::Measures::2D::MI - Perl module that provides error checks
                              for Loglieklihood, Total Mutual
                              Information, Pointwise Mutual Information
                              and Poisson-Stirling Measure.

SYNOPSIS

Basic Usage 
   use Text::NSP::Measures::2D::MI::ll;
 
 

   my $npp = 60; my $n1p = 20; my $np1 = 20;  my $n11 = 10;
 
 

   $ll_value = calculateStatistic( n11=>$n11,
                                       n1p=>$n1p,
                                       np1=>$np1,
                                       npp=>$npp);
 
 

   if( ($errorCode = getErrorCode()))
   {
     print STDERR $errorCode." - ".getErrorMessage()."\n"";
   }
   else
   {
     print getStatisticName."value for bigram is ".$ll_value."\n"";
   }

DESCRIPTION

This module is the base class for the Loglikelihood, Total Mutual Information and the Pointwise Mutual Information measures. All these measure are similar. This module provides error checks specific for these measures, it also implements the computations that are common to these measures.
Log-Likelihood measure is computed as
Log-Likelihood = 2 * [n11 * log(n11/m11) + n12 * log(n12/m12) +
                 n21 * log(n21/m21) + n22 * log(n22/m22)]
Total Mutual Information
TMI = (1/npp)*[n11 * log(n11/m11)/log 2 + n12 * log(n12/m12)/log 2 +
                 n21 * log(n21/m21)/log 2 + n22 * log(n22/m22)/log 2]
Pointwise Mutual Information
PMI = log (n11/m11)/log 2
Poisson Stirling Measures
PS = n11*(log (n11/m11)-1)

All these methods use the ratio of the observed values to expected values, for computations, and thus have common error checks, sothey have been grouped togrther.

Methods

getValues() - This method calls the computeMarginalTotals(), computeObservedValues() and the computeExpectedValues() methods to compute the observed and expected values. It checks these values for any errors that might cause the Loglikelihood, TMI & PMI measures to fail.
INPUT PARAMS : $count_values .. Reference of an hash containing
                                           the count values computed by the
                                           count.pl program.

RETURN VALUES : 1/undef ..returns '1' to indicate success
                                    and an undefined(NULL) value to indicate
                                    faliure.

computePMI() - Computes the pmi of a given observed and expected value pair.
INPUT PARAMS : $n ..Observed value
                $m         ..Expected value

RETURN VALUES : log(n/m) ..the log of the ratio of
                             observed value to expected
                             value.

AUTHOR

Ted Pedersen, University of Minnesota Duluth
                             <tpederse@d.umn.edu>

Satanjeev Banerjee, Carnegie Mellon University
                             <satanjeev@cmu.edu>

Amruta Purandare, University of Pittsburgh
                             <amruta@cs.pitt.edu>

Bridget Thomson-McInnes, University of Minnesota Twin Cities
                             <bthompson@d.umn.edu>

Saiyam Kohli, University of Minnesota Duluth
                             <kohli003@d.umn.edu>

HISTORY

Last updated: $Id: MI.pm,v 1.26 2006/09/15 23:27:03 saiyam_kohli Exp $

BUGS

SEE ALSO

<http://groups.yahoo.com/group/ngram/>

<http://www.d.umn.edu/~tpederse/nsp.html>

Copyright (C) 2000-2006, Ted Pedersen, Satanjeev Banerjee, Amruta Purandare, Bridget Thomson-McInnes and Saiyam Kohli

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.

Note: a copy of the GNU General Public License is available on the web at <http://www.gnu.org/licenses/gpl.txt> and is included in this distribution as GPL.txt.