Created
August 15, 2010 01:18
-
-
Save benui-dev/524931 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
use strict; | |
use warnings; | |
use Data::Dumper; | |
# Assume that we've got sentence-aligned stuff | |
my $sentence_pairs = [ | |
[ [ qw( the book ) ], [ qw( das buch ) ], ], | |
# [ [ qw( the house ) ], [ qw( das haus ) ], ], | |
# [ [ qw( a book ) ], [ qw( ein buch ) ], ], | |
# [ [ qw( a house ) ], [ qw( ein haus ) ], ], | |
# [ [ qw( a the book house ) ], [ qw( ein das buch haus ) ], ], | |
]; | |
# Shortcut, could have generated these | |
my @english_words = qw( the book ); | |
my @german_words = qw( das buch ); | |
#my @english_words = qw( the a book house ); | |
#my @german_words = qw( das ein buch haus ); | |
# Set up t uniformly | |
# t(e|f) | |
# cheat | |
my $t; | |
foreach my $e (@english_words) { | |
foreach my $g (@german_words) { | |
$t->{$g}->{$e} = 0.25; | |
} | |
} | |
# Should loop through until Expectation converges with Maximisation | |
# Can use perplexity to work out when we need to stop | |
foreach my $x ( 1...10 ) { | |
# Initialise counts to zero | |
# count(e|f) | |
my $count; | |
foreach my $g (@german_words) { | |
foreach my $e (@english_words) { | |
$count->{$g}->{$e} = 0; | |
} | |
} | |
# Initialize total to zero | |
# total(f) | |
my %gtotal; | |
foreach my $g (@german_words) { | |
$gtotal{$g} = 0; | |
} | |
# for all sentence pairs | |
foreach my $sentence_pair (@$sentence_pairs) { | |
my ($english, $german) = @$sentence_pair; | |
print "Training '" . join(' ', @$english) . "' and '" . join(' ', @$german) . "'\n"; | |
# Calculate total probability for each English word | |
# based on t(e|f) | |
my %etotal; | |
foreach my $e (@$english) { | |
$etotal{$e} = 0; | |
foreach my $g (@$german) { | |
$etotal{$e} += $t->{$g}->{$e}; | |
} | |
} | |
print "English total: " . Dumper(\%etotal); | |
# | |
foreach my $e (@$english) { | |
foreach my $g (@$german) { | |
print $count->{$g}->{$e} . " += " . $t->{$g}->{$e} . " / " . $etotal{$e} . "\n"; | |
$count->{$g}->{$e} += $t->{$g}->{$e} / $etotal{$e}; | |
$gtotal{$g} += $t->{$g}->{$e} / $etotal{$e}; | |
} | |
} | |
print "Count: " . Dumper($count); | |
print "gTotal: " . Dumper(\%gtotal); | |
} | |
# Estimate probabilities | |
# Based on the counts we've esen, update the probabilities t(e|f) | |
foreach my $g (@german_words) { | |
foreach my $e (@english_words) { | |
$t->{$g}->{$e} += $count->{$g}->{$e} / $gtotal{$g}; | |
} | |
} | |
print Dumper($t); | |
print "Hit Enter to continue..."; | |
my $meh = <stdin>; | |
} | |
# OK we're converged and done. Or somethin |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment