blib/lib/Algorithm/WordLevelStatistics.pm

filter: Statementstmt 100% 45 / 45 Branchbran 100% 4 / 4 Conditioncond n/a MC/DCmcdc n/a Subroutinesub 100% 6 / 6 Podpod 100% 3 / 3 totaltotal 100% 58 / 58 CC 6 SCAR 8.2
CC
6
Cov
100%
compute_spectrum
3.0
compute_spectra
3.0
_mean_and_variance
2.0
CRAP 2.3
↑ Summary
1 # Copyright 2009 Francesco Nidito. All rights reserved.
2 #
3 # This library is free software; you can redistribute it and/or
4 # modify it under the same terms as Perl itself.
5  
6 package Algorithm::WordLevelStatistics;
7  
8 1 ▶use strict;
Subroutine BEGIN: called CC=1 SCAR=0.0
9  
10 1 ▶use vars qw($VERSION);
Subroutine BEGIN: called CC=1 SCAR=0.0
11 1 $VERSION = '0.03';
12  
13 1 ▶sub new {
Subroutine new: called CC=1 SCAR=0.0
Pod documented
14 1 my $class = shift;
15 1 return bless {
16 version => $VERSION,
17 }, $class;
18 }
19  
20 # computes the statistical level of a single word (given its spectrum and the total number of words in the text)
21 2907 ▶sub compute_spectrum {
Subroutine compute_spectrum: called CC=3 SCAR=11.0
Pod documented
22 2907 my ($self, $N, $s) = @_;
23  
24 2907 my $n = @{$s};
25  
26 2907 my $ls = { count => $n, C => 0, sigma_nor => 0 };
27 2907 ▶ if( $n > 3 ) {
Branch: 1 branch
BranchTrueFalseTotal
if ($n > 3) 973 1934 2907
28 # position -> distance from preceding element in text
29 973 my @tmp = ();
30 973 ▶ for( my $i = 0; $i < ($n-1); ++$i ){ push @tmp, ($s->[$i+1] - $s->[$i]); }
Branch: 1 branch
BranchTrueFalseTotal
if ($i < $n - 1) 28079 973 29052
31  
32 973 my ($avg, $sigma) = $self->_mean_and_variance( \@tmp );
33 973 $sigma = sqrt($sigma)/$avg;
34  
35 # normalize sigma using an hypothetical uniform distribution
36 973 my $p = $n/$N;
37 973 $ls->{sigma_nor} = $sigma/sqrt(1.0-$p);
38  
39 # this is not simple:
40 973 $ls->{C} = ($ls->{sigma_nor} - (2.0*$n-1.0)/(2.0*$n+2.0)) * ( sqrt($n)*(1.0+2.8*$n**-0.865) );
41 }
42  
43 2907 return $ls;
44 }
45  
46 # computes the statistical level of a group of words (given their spectra)
47 1 ▶sub compute_spectra {
Subroutine compute_spectra: called CC=3 SCAR=11.0
Pod documented
48 1 my ($self, $s) = @_;
49  
50 # count the total number of "words" in text
51 1 my $N = 0;
52 1 foreach my $i (keys(%{$s})){ $N += @{ $s->{$i} }; }
53  
54 # computes the level statistic for all terms
55 1 my %r = ();
56 1 foreach my $i (keys(%{$s})){
57 2907 $r{$i} = $self->compute_spectrum( $N, $s->{$i} );
58 }
59  
60 1 return \%r;
61 }
62  
63 # fast, on-line algorithm to compute mean and variance:
64 # http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#On-line_algorithm
65 973 ▶sub _mean_and_variance {
Subroutine _mean_and_variance: called CC=2 SCAR=6.9
66 973 my ($self, $v) = @_;
67 973 my ($n, $mean, $M2) = (0, 0, 0);
68  
69 973 foreach my $x (@{$v}) {
70 28079 $n++;
71 28079 my $delta = $x - $mean;
72 28079 $mean += $delta/$n;
73 28079 $M2 += $delta*($x - $mean);
74 }
75  
76 973 my $variance = $M2/$n;
77  
78 973 return ($mean, $variance);
79 }
80 1 1;
81  
82 __END__
↑ Summary