line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package Treex::Block::W2A::ResegmentSentences; |
2
|
|
|
|
|
|
|
BEGIN { |
3
|
1
|
|
|
1
|
|
53413
|
$Treex::Block::W2A::ResegmentSentences::VERSION = '0.08170'; |
4
|
|
|
|
|
|
|
} |
5
|
1
|
|
|
1
|
|
2163
|
use Moose; |
|
0
|
|
|
|
|
|
|
|
0
|
|
|
|
|
|
|
6
|
|
|
|
|
|
|
use Treex::Core::Common; |
7
|
|
|
|
|
|
|
extends 'Treex::Core::Block'; |
8
|
|
|
|
|
|
|
|
9
|
|
|
|
|
|
|
has 'segmenters' => ( |
10
|
|
|
|
|
|
|
is => 'rw', |
11
|
|
|
|
|
|
|
isa => 'HashRef[Treex::Tool::Segment::RuleBased]', |
12
|
|
|
|
|
|
|
default => sub { return {} }, |
13
|
|
|
|
|
|
|
); |
14
|
|
|
|
|
|
|
|
15
|
|
|
|
|
|
|
sub _get_segmenter { |
16
|
|
|
|
|
|
|
my $self = shift; |
17
|
|
|
|
|
|
|
my $lang = uc shift; |
18
|
|
|
|
|
|
|
if ( exists $self->segmenters->{$lang} ) { |
19
|
|
|
|
|
|
|
return $self->segmenters->{$lang}; |
20
|
|
|
|
|
|
|
} |
21
|
|
|
|
|
|
|
my $specific = "Treex::Tool::Segment::${lang}::RuleBased"; |
22
|
|
|
|
|
|
|
my $fallback = "Treex::Tool::Segment::RuleBased"; |
23
|
|
|
|
|
|
|
foreach my $class ( $specific, $fallback ) { |
24
|
|
|
|
|
|
|
my $segmenter = eval "use $class; $class->new()"; ##no critic (BuiltinFunctions::ProhibitStringyEval) We want to use it, it is simpler and we check result |
25
|
|
|
|
|
|
|
if ($segmenter) { |
26
|
|
|
|
|
|
|
$self->segmenters->{$lang} = $segmenter; |
27
|
|
|
|
|
|
|
return $segmenter; |
28
|
|
|
|
|
|
|
} |
29
|
|
|
|
|
|
|
else { |
30
|
|
|
|
|
|
|
log_info("Failed during creating segmenter $class: $@"); |
31
|
|
|
|
|
|
|
} |
32
|
|
|
|
|
|
|
} |
33
|
|
|
|
|
|
|
log_fatal("Cannot create segmenter for $lang"); |
34
|
|
|
|
|
|
|
} |
35
|
|
|
|
|
|
|
|
36
|
|
|
|
|
|
|
sub process_bundle { |
37
|
|
|
|
|
|
|
my ( $self, $bundle ) = @_; |
38
|
|
|
|
|
|
|
|
39
|
|
|
|
|
|
|
my $my_label = $self->zone_label || ''; |
40
|
|
|
|
|
|
|
my %sentences; |
41
|
|
|
|
|
|
|
my ( $my_segments, $max_segments ) = ( 0, 0 ); |
42
|
|
|
|
|
|
|
foreach my $zone ( $bundle->get_all_zones() ) { |
43
|
|
|
|
|
|
|
my $lang = $zone->language; |
44
|
|
|
|
|
|
|
my $label = $zone->get_label(); |
45
|
|
|
|
|
|
|
my $segmenter = $self->_get_segmenter($lang); |
46
|
|
|
|
|
|
|
$sentences{$label} = [ $segmenter->get_segments( $zone->sentence ) ]; |
47
|
|
|
|
|
|
|
my $segments = @{ $sentences{$label} }; |
48
|
|
|
|
|
|
|
if ( $segments > $max_segments ) { $max_segments = $segments; } |
49
|
|
|
|
|
|
|
if ( $label eq $my_label ) { $my_segments = $segments; } |
50
|
|
|
|
|
|
|
} |
51
|
|
|
|
|
|
|
|
52
|
|
|
|
|
|
|
# If no language (and selector) were specified for this block |
53
|
|
|
|
|
|
|
# resegment all zones |
54
|
|
|
|
|
|
|
if ( $my_segments == 0 ) { |
55
|
|
|
|
|
|
|
$my_segments = $max_segments; |
56
|
|
|
|
|
|
|
} |
57
|
|
|
|
|
|
|
|
58
|
|
|
|
|
|
|
# We are finished if |
59
|
|
|
|
|
|
|
# the zone to be processed contains just one sentence. |
60
|
|
|
|
|
|
|
return if $my_segments == 1; |
61
|
|
|
|
|
|
|
|
62
|
|
|
|
|
|
|
# TODO: If a zone contains less subsegments (e.g. just 1) than $segments |
63
|
|
|
|
|
|
|
# we can try to split it to equally long chunks regardless of the real |
64
|
|
|
|
|
|
|
# sentence boundaries. Anyway, all evaluation blocks should join the |
65
|
|
|
|
|
|
|
# segments together again before measuring BLEU etc. |
66
|
|
|
|
|
|
|
my $doc = $bundle->get_document; |
67
|
|
|
|
|
|
|
my $orig_id = $bundle->id; |
68
|
|
|
|
|
|
|
$bundle->set_id("${orig_id}_1of$my_segments"); |
69
|
|
|
|
|
|
|
foreach my $zone ( $bundle->get_all_zones() ) { |
70
|
|
|
|
|
|
|
my $label = $zone->get_label(); |
71
|
|
|
|
|
|
|
my $sent = shift @{ $sentences{$label} }; |
72
|
|
|
|
|
|
|
$zone->set_sentence($sent); |
73
|
|
|
|
|
|
|
} |
74
|
|
|
|
|
|
|
my $last_bundle = $bundle; |
75
|
|
|
|
|
|
|
my @labels = keys %sentences; |
76
|
|
|
|
|
|
|
|
77
|
|
|
|
|
|
|
# TODO parameter to set how many bundles should be created: $my_segments or $max_segments? |
78
|
|
|
|
|
|
|
for my $i ( 2 .. $my_segments ) { |
79
|
|
|
|
|
|
|
my $new_bundle = $doc->create_bundle( { after => $last_bundle } ); |
80
|
|
|
|
|
|
|
$last_bundle = $new_bundle; |
81
|
|
|
|
|
|
|
$new_bundle->set_id("${orig_id}_${i}of$my_segments"); |
82
|
|
|
|
|
|
|
foreach my $label (@labels) { |
83
|
|
|
|
|
|
|
my $sent = shift @{ $sentences{$label} }; |
84
|
|
|
|
|
|
|
if ( !defined $sent ) { $sent = ' '; } |
85
|
|
|
|
|
|
|
|
86
|
|
|
|
|
|
|
# If some zone contains more segments than the "current" zone, |
87
|
|
|
|
|
|
|
# the remaining segments will be joined to the last bundle. |
88
|
|
|
|
|
|
|
if ( $i == $my_segments && $max_segments > $my_segments ) { |
89
|
|
|
|
|
|
|
$sent .= ' ' . join( ' ', @{ $sentences{$label} } ); |
90
|
|
|
|
|
|
|
} |
91
|
|
|
|
|
|
|
my ( $lang, $selector ) = split /_/, $label; |
92
|
|
|
|
|
|
|
my $new_zone = $new_bundle->create_zone( $lang, $selector ); |
93
|
|
|
|
|
|
|
$new_zone->set_sentence($sent); |
94
|
|
|
|
|
|
|
} |
95
|
|
|
|
|
|
|
} |
96
|
|
|
|
|
|
|
|
97
|
|
|
|
|
|
|
return; |
98
|
|
|
|
|
|
|
} |
99
|
|
|
|
|
|
|
|
100
|
|
|
|
|
|
|
1; |
101
|
|
|
|
|
|
|
|
102
|
|
|
|
|
|
|
__END__ |
103
|
|
|
|
|
|
|
|
104
|
|
|
|
|
|
|
TODO POD |
105
|
|
|
|
|
|
|
|
106
|
|
|
|
|
|
|
=encoding utf-8 |
107
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
=head1 NAME |
109
|
|
|
|
|
|
|
|
110
|
|
|
|
|
|
|
Treex::Block::W2A::ResegmentSentences |
111
|
|
|
|
|
|
|
|
112
|
|
|
|
|
|
|
=head1 VERSION |
113
|
|
|
|
|
|
|
|
114
|
|
|
|
|
|
|
version 0.08170 |
115
|
|
|
|
|
|
|
|
116
|
|
|
|
|
|
|
=head1 DESCRIPTION |
117
|
|
|
|
|
|
|
|
118
|
|
|
|
|
|
|
If the sentence segmenter says that the current sentence is |
119
|
|
|
|
|
|
|
actually composed of two or more sentences, then new bundles |
120
|
|
|
|
|
|
|
are inserted after the current bundle, each containing just |
121
|
|
|
|
|
|
|
one piece of the resegmented original sentence. |
122
|
|
|
|
|
|
|
|
123
|
|
|
|
|
|
|
All zones are processed. |
124
|
|
|
|
|
|
|
The number of bundles created is determined by the number of subsegments |
125
|
|
|
|
|
|
|
in the "current" zone (specified by the parameters C<language> and C<selector>). |
126
|
|
|
|
|
|
|
If a zone contains less subsegments than the current one, |
127
|
|
|
|
|
|
|
the remaining bundles will contain empty sentence. |
128
|
|
|
|
|
|
|
If a zone contains more subsegments than the current one, |
129
|
|
|
|
|
|
|
the remaining subsegments will be joined in the last bundle. |
130
|
|
|
|
|
|
|
|
131
|
|
|
|
|
|
|
In other words, it is granted that the current zone, |
132
|
|
|
|
|
|
|
will not contain empty sentences. |
133
|
|
|
|
|
|
|
|
134
|
|
|
|
|
|
|
=head1 AUTHOR |
135
|
|
|
|
|
|
|
|
136
|
|
|
|
|
|
|
ZdenÄk Žabokrtský <zabokrtsky@ufal.mff.cuni.cz> |
137
|
|
|
|
|
|
|
|
138
|
|
|
|
|
|
|
Martin Popel <popel@ufal.mff.cuni.cz> |
139
|
|
|
|
|
|
|
|
140
|
|
|
|
|
|
|
=head1 COPYRIGHT AND LICENSE |
141
|
|
|
|
|
|
|
|
142
|
|
|
|
|
|
|
Copyright © 2011 by Institute of Formal and Applied Linguistics, Charles University in Prague |
143
|
|
|
|
|
|
|
|
144
|
|
|
|
|
|
|
This module is free software; you can redistribute it and/or modify it under the same terms as Perl itself. |