line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package AI::Categorizer::KnowledgeSet; |
2
|
|
|
|
|
|
|
|
3
|
11
|
|
|
11
|
|
77
|
use strict; |
|
11
|
|
|
|
|
21
|
|
|
11
|
|
|
|
|
388
|
|
4
|
11
|
|
|
11
|
|
52
|
use Class::Container; |
|
11
|
|
|
|
|
17
|
|
|
11
|
|
|
|
|
183
|
|
5
|
11
|
|
|
11
|
|
51
|
use AI::Categorizer::Storable; |
|
11
|
|
|
|
|
18
|
|
|
11
|
|
|
|
|
238
|
|
6
|
11
|
|
|
11
|
|
45
|
use base qw(Class::Container AI::Categorizer::Storable); |
|
11
|
|
|
|
|
19
|
|
|
11
|
|
|
|
|
972
|
|
7
|
|
|
|
|
|
|
|
8
|
11
|
|
|
11
|
|
50
|
use Params::Validate qw(:types); |
|
11
|
|
|
|
|
19
|
|
|
11
|
|
|
|
|
1547
|
|
9
|
11
|
|
|
11
|
|
52
|
use AI::Categorizer::ObjectSet; |
|
11
|
|
|
|
|
17
|
|
|
11
|
|
|
|
|
215
|
|
10
|
11
|
|
|
11
|
|
46
|
use AI::Categorizer::Document; |
|
11
|
|
|
|
|
21
|
|
|
11
|
|
|
|
|
249
|
|
11
|
11
|
|
|
11
|
|
48
|
use AI::Categorizer::Category; |
|
11
|
|
|
|
|
22
|
|
|
11
|
|
|
|
|
285
|
|
12
|
11
|
|
|
11
|
|
50
|
use AI::Categorizer::FeatureVector; |
|
11
|
|
|
|
|
23
|
|
|
11
|
|
|
|
|
199
|
|
13
|
11
|
|
|
11
|
|
5363
|
use AI::Categorizer::Util; |
|
11
|
|
|
|
|
26
|
|
|
11
|
|
|
|
|
530
|
|
14
|
11
|
|
|
11
|
|
59
|
use Carp qw(croak); |
|
11
|
|
|
|
|
23
|
|
|
11
|
|
|
|
|
32497
|
|
15
|
|
|
|
|
|
|
|
16
|
|
|
|
|
|
|
__PACKAGE__->valid_params |
17
|
|
|
|
|
|
|
( |
18
|
|
|
|
|
|
|
categories => { |
19
|
|
|
|
|
|
|
type => ARRAYREF, |
20
|
|
|
|
|
|
|
default => [], |
21
|
|
|
|
|
|
|
callbacks => { 'all are Category objects' => |
22
|
|
|
|
|
|
|
sub { ! grep !UNIVERSAL::isa($_, 'AI::Categorizer::Category'), |
23
|
|
|
|
|
|
|
@{$_[0]} }, |
24
|
|
|
|
|
|
|
}, |
25
|
|
|
|
|
|
|
}, |
26
|
|
|
|
|
|
|
documents => { |
27
|
|
|
|
|
|
|
type => ARRAYREF, |
28
|
|
|
|
|
|
|
default => [], |
29
|
|
|
|
|
|
|
callbacks => { 'all are Document objects' => |
30
|
|
|
|
|
|
|
sub { ! grep !UNIVERSAL::isa($_, 'AI::Categorizer::Document'), |
31
|
|
|
|
|
|
|
@{$_[0]} }, |
32
|
|
|
|
|
|
|
}, |
33
|
|
|
|
|
|
|
}, |
34
|
|
|
|
|
|
|
scan_first => { |
35
|
|
|
|
|
|
|
type => BOOLEAN, |
36
|
|
|
|
|
|
|
default => 1, |
37
|
|
|
|
|
|
|
}, |
38
|
|
|
|
|
|
|
feature_selector => { |
39
|
|
|
|
|
|
|
isa => 'AI::Categorizer::FeatureSelector', |
40
|
|
|
|
|
|
|
}, |
41
|
|
|
|
|
|
|
tfidf_weighting => { |
42
|
|
|
|
|
|
|
type => SCALAR, |
43
|
|
|
|
|
|
|
optional => 1, |
44
|
|
|
|
|
|
|
}, |
45
|
|
|
|
|
|
|
term_weighting => { |
46
|
|
|
|
|
|
|
type => SCALAR, |
47
|
|
|
|
|
|
|
default => 'x', |
48
|
|
|
|
|
|
|
}, |
49
|
|
|
|
|
|
|
collection_weighting => { |
50
|
|
|
|
|
|
|
type => SCALAR, |
51
|
|
|
|
|
|
|
default => 'x', |
52
|
|
|
|
|
|
|
}, |
53
|
|
|
|
|
|
|
normalize_weighting => { |
54
|
|
|
|
|
|
|
type => SCALAR, |
55
|
|
|
|
|
|
|
default => 'x', |
56
|
|
|
|
|
|
|
}, |
57
|
|
|
|
|
|
|
verbose => { |
58
|
|
|
|
|
|
|
type => SCALAR, |
59
|
|
|
|
|
|
|
default => 0, |
60
|
|
|
|
|
|
|
}, |
61
|
|
|
|
|
|
|
); |
62
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
__PACKAGE__->contained_objects |
64
|
|
|
|
|
|
|
( |
65
|
|
|
|
|
|
|
document => { delayed => 1, |
66
|
|
|
|
|
|
|
class => 'AI::Categorizer::Document' }, |
67
|
|
|
|
|
|
|
category => { delayed => 1, |
68
|
|
|
|
|
|
|
class => 'AI::Categorizer::Category' }, |
69
|
|
|
|
|
|
|
collection => { delayed => 1, |
70
|
|
|
|
|
|
|
class => 'AI::Categorizer::Collection::Files' }, |
71
|
|
|
|
|
|
|
features => { delayed => 1, |
72
|
|
|
|
|
|
|
class => 'AI::Categorizer::FeatureVector' }, |
73
|
|
|
|
|
|
|
feature_selector => 'AI::Categorizer::FeatureSelector::DocFrequency', |
74
|
|
|
|
|
|
|
); |
75
|
|
|
|
|
|
|
|
76
|
|
|
|
|
|
|
sub new { |
77
|
12
|
|
|
12
|
1
|
6934
|
my ($pkg, %args) = @_; |
78
|
|
|
|
|
|
|
|
79
|
|
|
|
|
|
|
# Shortcuts |
80
|
12
|
100
|
|
|
|
52
|
if ($args{tfidf_weighting}) { |
81
|
1
|
|
|
|
|
5
|
@args{'term_weighting', 'collection_weighting', 'normalize_weighting'} = split '', $args{tfidf_weighting}; |
82
|
1
|
|
|
|
|
4
|
delete $args{tfidf_weighting}; |
83
|
|
|
|
|
|
|
} |
84
|
|
|
|
|
|
|
|
85
|
12
|
|
|
|
|
133
|
my $self = $pkg->SUPER::new(%args); |
86
|
|
|
|
|
|
|
|
87
|
|
|
|
|
|
|
# Convert to AI::Categorizer::ObjectSet sets |
88
|
12
|
|
|
|
|
8460
|
$self->{categories} = new AI::Categorizer::ObjectSet( @{$self->{categories}} ); |
|
12
|
|
|
|
|
94
|
|
89
|
12
|
|
|
|
|
24
|
$self->{documents} = new AI::Categorizer::ObjectSet( @{$self->{documents}} ); |
|
12
|
|
|
|
|
47
|
|
90
|
|
|
|
|
|
|
|
91
|
12
|
50
|
|
|
|
58
|
if ($self->{load}) { |
92
|
0
|
0
|
|
|
|
0
|
my $args = ref($self->{load}) ? $self->{load} : { path => $self->{load} }; |
93
|
0
|
|
|
|
|
0
|
$self->load(%$args); |
94
|
0
|
|
|
|
|
0
|
delete $self->{load}; |
95
|
|
|
|
|
|
|
} |
96
|
12
|
|
|
|
|
56
|
return $self; |
97
|
|
|
|
|
|
|
} |
98
|
|
|
|
|
|
|
|
99
|
|
|
|
|
|
|
sub features { |
100
|
19
|
|
|
19
|
1
|
37
|
my $self = shift; |
101
|
|
|
|
|
|
|
|
102
|
19
|
100
|
|
|
|
41
|
if (@_) { |
103
|
1
|
|
|
|
|
2
|
$self->{features} = shift; |
104
|
1
|
50
|
|
|
|
5
|
$self->trim_doc_features if $self->{features}; |
105
|
|
|
|
|
|
|
} |
106
|
19
|
100
|
|
|
|
88
|
return $self->{features} if $self->{features}; |
107
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
# Create a feature vector encompassing the whole set of documents |
109
|
3
|
|
|
|
|
12
|
my $v = $self->create_delayed_object('features'); |
110
|
3
|
|
|
|
|
10
|
foreach my $document ($self->documents) { |
111
|
12
|
|
|
|
|
35
|
$v->add( $document->features ); |
112
|
|
|
|
|
|
|
} |
113
|
3
|
|
|
|
|
15
|
return $self->{features} = $v; |
114
|
|
|
|
|
|
|
} |
115
|
|
|
|
|
|
|
|
116
|
|
|
|
|
|
|
sub categories { |
117
|
24
|
|
|
24
|
1
|
43
|
my $c = $_[0]->{categories}; |
118
|
24
|
50
|
|
|
|
97
|
return wantarray ? $c->members : $c->size; |
119
|
|
|
|
|
|
|
} |
120
|
|
|
|
|
|
|
|
121
|
|
|
|
|
|
|
sub documents { |
122
|
35
|
|
|
35
|
1
|
67
|
my $d = $_[0]->{documents}; |
123
|
35
|
100
|
|
|
|
127
|
return wantarray ? $d->members : $d->size; |
124
|
|
|
|
|
|
|
} |
125
|
|
|
|
|
|
|
|
126
|
|
|
|
|
|
|
sub document { |
127
|
7
|
|
|
7
|
1
|
14
|
my ($self, $name) = @_; |
128
|
7
|
|
|
|
|
24
|
return $self->{documents}->retrieve($name); |
129
|
|
|
|
|
|
|
} |
130
|
|
|
|
|
|
|
|
131
|
0
|
|
|
0
|
0
|
0
|
sub feature_selector { $_[0]->{feature_selector} } |
132
|
0
|
|
|
0
|
0
|
0
|
sub scan_first { $_[0]->{scan_first} } |
133
|
|
|
|
|
|
|
|
134
|
|
|
|
|
|
|
sub verbose { |
135
|
0
|
|
|
0
|
1
|
0
|
my $self = shift; |
136
|
0
|
0
|
|
|
|
0
|
$self->{verbose} = shift if @_; |
137
|
0
|
|
|
|
|
0
|
return $self->{verbose}; |
138
|
|
|
|
|
|
|
} |
139
|
|
|
|
|
|
|
|
140
|
|
|
|
|
|
|
sub trim_doc_features { |
141
|
0
|
|
|
0
|
0
|
0
|
my ($self) = @_; |
142
|
|
|
|
|
|
|
|
143
|
0
|
|
|
|
|
0
|
foreach my $doc ($self->documents) { |
144
|
0
|
|
|
|
|
0
|
$doc->features( $doc->features->intersection($self->features) ); |
145
|
|
|
|
|
|
|
} |
146
|
|
|
|
|
|
|
} |
147
|
|
|
|
|
|
|
|
148
|
|
|
|
|
|
|
|
149
|
|
|
|
|
|
|
sub prog_bar { |
150
|
0
|
|
|
0
|
0
|
0
|
my ($self, $collection) = @_; |
151
|
|
|
|
|
|
|
|
152
|
0
|
0
|
|
0
|
|
0
|
return sub {} unless $self->verbose; |
|
0
|
|
|
|
|
0
|
|
153
|
0
|
0
|
|
0
|
|
0
|
return sub { print STDERR '.' } unless eval "use Time::Progress; 1"; |
|
0
|
|
|
|
|
0
|
|
154
|
|
|
|
|
|
|
|
155
|
0
|
0
|
|
|
|
0
|
my $count = $collection->can('count_documents') ? $collection->count_documents : 0; |
156
|
|
|
|
|
|
|
|
157
|
0
|
|
|
|
|
0
|
my $pb = 'Time::Progress'->new; |
158
|
0
|
|
|
|
|
0
|
$pb->attr(max => $count); |
159
|
0
|
|
|
|
|
0
|
my $i = 0; |
160
|
|
|
|
|
|
|
return sub { |
161
|
0
|
|
|
0
|
|
0
|
$i++; |
162
|
0
|
0
|
|
|
|
0
|
return if $i % 25; |
163
|
0
|
|
|
|
|
0
|
print STDERR $pb->report("%50b %p ($i/$count)\r", $i); |
164
|
0
|
|
|
|
|
0
|
}; |
165
|
|
|
|
|
|
|
} |
166
|
|
|
|
|
|
|
|
167
|
|
|
|
|
|
|
# A little utility method for several other methods like scan_stats(), |
168
|
|
|
|
|
|
|
# load(), read(), etc. |
169
|
|
|
|
|
|
|
sub _make_collection { |
170
|
0
|
|
|
0
|
|
0
|
my ($self, $args) = @_; |
171
|
0
|
|
0
|
|
|
0
|
return $args->{collection} || $self->create_delayed_object('collection', %$args); |
172
|
|
|
|
|
|
|
} |
173
|
|
|
|
|
|
|
|
174
|
|
|
|
|
|
|
sub scan_stats { |
175
|
|
|
|
|
|
|
# Should determine: |
176
|
|
|
|
|
|
|
# - number of documents |
177
|
|
|
|
|
|
|
# - number of categories |
178
|
|
|
|
|
|
|
# - avg. number of categories per document (whole corpus) |
179
|
|
|
|
|
|
|
# - avg. number of tokens per document (whole corpus) |
180
|
|
|
|
|
|
|
# - avg. number of types per document (whole corpus) |
181
|
|
|
|
|
|
|
# - number of documents, tokens, & types for each category |
182
|
|
|
|
|
|
|
# - "category skew index" (% variance?) by num. documents, tokens, and types |
183
|
|
|
|
|
|
|
|
184
|
0
|
|
|
0
|
1
|
0
|
my ($self, %args) = @_; |
185
|
0
|
|
|
|
|
0
|
my $collection = $self->_make_collection(\%args); |
186
|
0
|
|
|
|
|
0
|
my $pb = $self->prog_bar($collection); |
187
|
|
|
|
|
|
|
|
188
|
0
|
|
|
|
|
0
|
my %stats; |
189
|
|
|
|
|
|
|
|
190
|
|
|
|
|
|
|
|
191
|
0
|
|
|
|
|
0
|
while (my $doc = $collection->next) { |
192
|
0
|
|
|
|
|
0
|
$pb->(); |
193
|
0
|
|
|
|
|
0
|
$stats{category_count_with_duplicates} += $doc->categories; |
194
|
|
|
|
|
|
|
|
195
|
0
|
|
|
|
|
0
|
my ($sum, $length) = ($doc->features->sum, $doc->features->length); |
196
|
0
|
|
|
|
|
0
|
$stats{document_count}++; |
197
|
0
|
|
|
|
|
0
|
$stats{token_count} += $sum; |
198
|
0
|
|
|
|
|
0
|
$stats{type_count} += $length; |
199
|
|
|
|
|
|
|
|
200
|
0
|
|
|
|
|
0
|
foreach my $cat ($doc->categories) { |
201
|
|
|
|
|
|
|
#warn $doc->name, ": ", $cat->name, "\n"; |
202
|
0
|
|
|
|
|
0
|
$stats{categories}{$cat->name}{document_count}++; |
203
|
0
|
|
|
|
|
0
|
$stats{categories}{$cat->name}{token_count} += $sum; |
204
|
0
|
|
|
|
|
0
|
$stats{categories}{$cat->name}{type_count} += $length; |
205
|
|
|
|
|
|
|
} |
206
|
|
|
|
|
|
|
} |
207
|
0
|
0
|
|
|
|
0
|
print "\n" if $self->verbose; |
208
|
|
|
|
|
|
|
|
209
|
0
|
|
|
|
|
0
|
my @cats = keys %{ $stats{categories} }; |
|
0
|
|
|
|
|
0
|
|
210
|
|
|
|
|
|
|
|
211
|
0
|
|
|
|
|
0
|
$stats{category_count} = @cats; |
212
|
0
|
|
|
|
|
0
|
$stats{categories_per_document} = $stats{category_count_with_duplicates} / $stats{document_count}; |
213
|
0
|
|
|
|
|
0
|
$stats{tokens_per_document} = $stats{token_count} / $stats{document_count}; |
214
|
0
|
|
|
|
|
0
|
$stats{types_per_document} = $stats{type_count} / $stats{document_count}; |
215
|
|
|
|
|
|
|
|
216
|
0
|
|
|
|
|
0
|
foreach my $thing ('type', 'token', 'document') { |
217
|
0
|
|
|
|
|
0
|
$stats{"${thing}s_per_category"} = AI::Categorizer::Util::average |
218
|
0
|
|
|
|
|
0
|
( map { $stats{categories}{$_}{"${thing}_count"} } @cats ); |
219
|
|
|
|
|
|
|
|
220
|
0
|
0
|
|
|
|
0
|
next unless @cats; |
221
|
|
|
|
|
|
|
|
222
|
|
|
|
|
|
|
# Compute the skews |
223
|
0
|
|
|
|
|
0
|
my $ssum; |
224
|
0
|
|
|
|
|
0
|
foreach my $cat (@cats) { |
225
|
0
|
|
|
|
|
0
|
$ssum += ($stats{categories}{$cat}{"${thing}_count"} - $stats{"${thing}s_per_category"}) ** 2; |
226
|
|
|
|
|
|
|
} |
227
|
0
|
|
|
|
|
0
|
$stats{"${thing}_skew_by_category"} = sqrt($ssum/@cats) / $stats{"${thing}s_per_category"}; |
228
|
|
|
|
|
|
|
} |
229
|
|
|
|
|
|
|
|
230
|
0
|
|
|
|
|
0
|
return \%stats; |
231
|
|
|
|
|
|
|
} |
232
|
|
|
|
|
|
|
|
233
|
|
|
|
|
|
|
sub load { |
234
|
0
|
|
|
0
|
1
|
0
|
my ($self, %args) = @_; |
235
|
0
|
|
|
|
|
0
|
my $c = $self->_make_collection(\%args); |
236
|
|
|
|
|
|
|
|
237
|
0
|
0
|
|
|
|
0
|
if ($self->{features_kept}) { |
|
|
0
|
|
|
|
|
|
238
|
|
|
|
|
|
|
# Read the whole thing in, then reduce |
239
|
0
|
|
|
|
|
0
|
$self->read( collection => $c ); |
240
|
0
|
|
|
|
|
0
|
$self->select_features; |
241
|
|
|
|
|
|
|
|
242
|
|
|
|
|
|
|
} elsif ($self->{scan_first}) { |
243
|
|
|
|
|
|
|
# Figure out the feature set first, then read data in |
244
|
0
|
|
|
|
|
0
|
$self->scan_features( collection => $c ); |
245
|
0
|
|
|
|
|
0
|
$c->rewind; |
246
|
0
|
|
|
|
|
0
|
$self->read( collection => $c ); |
247
|
|
|
|
|
|
|
|
248
|
|
|
|
|
|
|
} else { |
249
|
|
|
|
|
|
|
# Don't do any feature reduction, just read the data |
250
|
0
|
|
|
|
|
0
|
$self->read( collection => $c ); |
251
|
|
|
|
|
|
|
} |
252
|
|
|
|
|
|
|
} |
253
|
|
|
|
|
|
|
|
254
|
|
|
|
|
|
|
sub read { |
255
|
0
|
|
|
0
|
1
|
0
|
my ($self, %args) = @_; |
256
|
0
|
|
|
|
|
0
|
my $collection = $self->_make_collection(\%args); |
257
|
0
|
|
|
|
|
0
|
my $pb = $self->prog_bar($collection); |
258
|
|
|
|
|
|
|
|
259
|
0
|
|
|
|
|
0
|
while (my $doc = $collection->next) { |
260
|
0
|
|
|
|
|
0
|
$pb->(); |
261
|
0
|
|
|
|
|
0
|
$self->add_document($doc); |
262
|
|
|
|
|
|
|
} |
263
|
0
|
0
|
|
|
|
0
|
print "\n" if $self->verbose; |
264
|
|
|
|
|
|
|
} |
265
|
|
|
|
|
|
|
|
266
|
|
|
|
|
|
|
sub finish { |
267
|
11
|
|
|
11
|
1
|
29
|
my $self = shift; |
268
|
11
|
100
|
|
|
|
51
|
return if $self->{finished}++; |
269
|
10
|
|
|
|
|
43
|
$self->weigh_features; |
270
|
|
|
|
|
|
|
} |
271
|
|
|
|
|
|
|
|
272
|
|
|
|
|
|
|
sub weigh_features { |
273
|
|
|
|
|
|
|
# This could be made more efficient by figuring out an execution |
274
|
|
|
|
|
|
|
# plan in advance |
275
|
|
|
|
|
|
|
|
276
|
10
|
|
|
10
|
1
|
16
|
my $self = shift; |
277
|
|
|
|
|
|
|
|
278
|
10
|
100
|
|
|
|
92
|
if ( $self->{term_weighting} =~ /^(t|x)$/ ) { |
|
|
50
|
|
|
|
|
|
|
|
100
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
279
|
|
|
|
|
|
|
# Nothing to do |
280
|
|
|
|
|
|
|
} elsif ( $self->{term_weighting} eq 'l' ) { |
281
|
0
|
|
|
|
|
0
|
foreach my $doc ($self->documents) { |
282
|
0
|
|
|
|
|
0
|
my $f = $doc->features->as_hash; |
283
|
0
|
|
|
|
|
0
|
$_ = 1 + log($_) foreach values %$f; |
284
|
|
|
|
|
|
|
} |
285
|
|
|
|
|
|
|
} elsif ( $self->{term_weighting} eq 'n' ) { |
286
|
1
|
|
|
|
|
6
|
foreach my $doc ($self->documents) { |
287
|
4
|
|
|
|
|
13
|
my $f = $doc->features->as_hash; |
288
|
4
|
|
|
|
|
19
|
my $max_tf = AI::Categorizer::Util::max values %$f; |
289
|
4
|
|
|
|
|
35
|
$_ = 0.5 + 0.5 * $_ / $max_tf foreach values %$f; |
290
|
|
|
|
|
|
|
} |
291
|
|
|
|
|
|
|
} elsif ( $self->{term_weighting} eq 'b' ) { |
292
|
1
|
|
|
|
|
6
|
foreach my $doc ($self->documents) { |
293
|
4
|
|
|
|
|
10
|
my $f = $doc->features->as_hash; |
294
|
4
|
50
|
|
|
|
37
|
$_ = $_ ? 1 : 0 foreach values %$f; |
295
|
|
|
|
|
|
|
} |
296
|
|
|
|
|
|
|
} else { |
297
|
0
|
|
|
|
|
0
|
die "term_weighting must be one of 'x', 't', 'l', 'b', or 'n'"; |
298
|
|
|
|
|
|
|
} |
299
|
|
|
|
|
|
|
|
300
|
10
|
100
|
|
|
|
53
|
if ($self->{collection_weighting} eq 'x') { |
|
|
50
|
|
|
|
|
|
301
|
|
|
|
|
|
|
# Nothing to do |
302
|
|
|
|
|
|
|
} elsif ($self->{collection_weighting} =~ /^(f|p)$/) { |
303
|
1
|
50
|
|
|
|
6
|
my $subtrahend = ($1 eq 'f' ? 0 : 1); |
304
|
1
|
|
|
|
|
5
|
my $num_docs = $self->documents; |
305
|
1
|
|
|
|
|
5
|
$self->document_frequency('foo'); # Initialize |
306
|
1
|
|
|
|
|
3
|
foreach my $doc ($self->documents) { |
307
|
4
|
|
|
|
|
9
|
my $f = $doc->features->as_hash; |
308
|
4
|
|
|
|
|
52
|
$f->{$_} *= log($num_docs / $self->{doc_freq_vector}{$_} - $subtrahend) foreach keys %$f; |
309
|
|
|
|
|
|
|
} |
310
|
|
|
|
|
|
|
} else { |
311
|
0
|
|
|
|
|
0
|
die "collection_weighting must be one of 'x', 'f', or 'p'"; |
312
|
|
|
|
|
|
|
} |
313
|
|
|
|
|
|
|
|
314
|
10
|
50
|
|
|
|
49
|
if ( $self->{normalize_weighting} eq 'x' ) { |
|
|
0
|
|
|
|
|
|
315
|
|
|
|
|
|
|
# Nothing to do |
316
|
|
|
|
|
|
|
} elsif ( $self->{normalize_weighting} eq 'c' ) { |
317
|
0
|
|
|
|
|
0
|
$_->features->normalize foreach $self->documents; |
318
|
|
|
|
|
|
|
} else { |
319
|
0
|
|
|
|
|
0
|
die "normalize_weighting must be one of 'x' or 'c'"; |
320
|
|
|
|
|
|
|
} |
321
|
|
|
|
|
|
|
} |
322
|
|
|
|
|
|
|
|
323
|
|
|
|
|
|
|
sub document_frequency { |
324
|
4
|
|
|
4
|
1
|
7
|
my ($self, $term) = @_; |
325
|
|
|
|
|
|
|
|
326
|
4
|
100
|
|
|
|
12
|
unless (exists $self->{doc_freq_vector}) { |
327
|
1
|
50
|
|
|
|
5
|
die "No corpus has been scanned for features" unless $self->documents; |
328
|
|
|
|
|
|
|
|
329
|
1
|
|
|
|
|
4
|
my $doc_freq = $self->create_delayed_object('features', features => {}); |
330
|
1
|
|
|
|
|
4
|
foreach my $doc ($self->documents) { |
331
|
4
|
|
|
|
|
11
|
$doc_freq->add( $doc->features->as_boolean_hash ); |
332
|
|
|
|
|
|
|
} |
333
|
1
|
|
|
|
|
5
|
$self->{doc_freq_vector} = $doc_freq->as_hash; |
334
|
|
|
|
|
|
|
} |
335
|
|
|
|
|
|
|
|
336
|
4
|
100
|
|
|
|
22
|
return exists $self->{doc_freq_vector}{$term} ? $self->{doc_freq_vector}{$term} : 0; |
337
|
|
|
|
|
|
|
} |
338
|
|
|
|
|
|
|
|
339
|
|
|
|
|
|
|
sub scan_features { |
340
|
0
|
|
|
0
|
1
|
0
|
my ($self, %args) = @_; |
341
|
0
|
|
|
|
|
0
|
my $c = $self->_make_collection(\%args); |
342
|
|
|
|
|
|
|
|
343
|
0
|
|
|
|
|
0
|
my $pb = $self->prog_bar($c); |
344
|
0
|
|
|
|
|
0
|
my $ranked_features = $self->{feature_selector}->scan_features( collection => $c, prog_bar => $pb ); |
345
|
|
|
|
|
|
|
|
346
|
0
|
|
|
|
|
0
|
$self->delayed_object_params('document', use_features => $ranked_features); |
347
|
0
|
|
|
|
|
0
|
$self->delayed_object_params('collection', use_features => $ranked_features); |
348
|
0
|
|
|
|
|
0
|
return $ranked_features; |
349
|
|
|
|
|
|
|
} |
350
|
|
|
|
|
|
|
|
351
|
|
|
|
|
|
|
sub select_features { |
352
|
0
|
|
|
0
|
0
|
0
|
my $self = shift; |
353
|
|
|
|
|
|
|
|
354
|
0
|
|
|
|
|
0
|
my $f = $self->feature_selector->select_features(knowledge_set => $self); |
355
|
0
|
|
|
|
|
0
|
$self->features($f); |
356
|
|
|
|
|
|
|
} |
357
|
|
|
|
|
|
|
|
358
|
|
|
|
|
|
|
sub partition { |
359
|
0
|
|
|
0
|
1
|
0
|
my ($self, @sizes) = @_; |
360
|
0
|
|
|
|
|
0
|
my $num_docs = my @docs = $self->documents; |
361
|
0
|
|
|
|
|
0
|
my @groups; |
362
|
|
|
|
|
|
|
|
363
|
0
|
|
|
|
|
0
|
while (@sizes > 1) { |
364
|
0
|
|
|
|
|
0
|
my $size = int ($num_docs * shift @sizes); |
365
|
0
|
|
|
|
|
0
|
push @groups, []; |
366
|
0
|
|
|
|
|
0
|
for (0..$size) { |
367
|
0
|
|
|
|
|
0
|
push @{ $groups[-1] }, splice @docs, rand(@docs), 1; |
|
0
|
|
|
|
|
0
|
|
368
|
|
|
|
|
|
|
} |
369
|
|
|
|
|
|
|
} |
370
|
0
|
|
|
|
|
0
|
push @groups, \@docs; |
371
|
|
|
|
|
|
|
|
372
|
0
|
|
|
|
|
0
|
return map { ref($self)->new( documents => $_ ) } @groups; |
|
0
|
|
|
|
|
0
|
|
373
|
|
|
|
|
|
|
} |
374
|
|
|
|
|
|
|
|
375
|
|
|
|
|
|
|
sub make_document { |
376
|
40
|
|
|
40
|
1
|
134
|
my ($self, %args) = @_; |
377
|
40
|
|
|
|
|
73
|
my $cats = delete $args{categories}; |
378
|
40
|
|
|
|
|
74
|
my @cats = map { $self->call_method('category', 'by_name', name => $_) } @$cats; |
|
40
|
|
|
|
|
176
|
|
379
|
40
|
|
|
|
|
168
|
my $d = $self->create_delayed_object('document', %args, categories => \@cats); |
380
|
40
|
|
|
|
|
126
|
$self->add_document($d); |
381
|
|
|
|
|
|
|
} |
382
|
|
|
|
|
|
|
|
383
|
|
|
|
|
|
|
sub add_document { |
384
|
40
|
|
|
40
|
1
|
55
|
my ($self, $doc) = @_; |
385
|
|
|
|
|
|
|
|
386
|
40
|
|
|
|
|
104
|
foreach ($doc->categories) { |
387
|
40
|
|
|
|
|
115
|
$_->add_document($doc); |
388
|
|
|
|
|
|
|
} |
389
|
40
|
|
|
|
|
133
|
$self->{documents}->insert($doc); |
390
|
40
|
|
|
|
|
117
|
$self->{categories}->insert($doc->categories); |
391
|
|
|
|
|
|
|
} |
392
|
|
|
|
|
|
|
|
393
|
|
|
|
|
|
|
sub save_features { |
394
|
0
|
|
|
0
|
1
|
|
my ($self, $file) = @_; |
395
|
|
|
|
|
|
|
|
396
|
0
|
0
|
0
|
|
|
|
my $f = ($self->{features} || { $self->delayed_object_params('document') }->{use_features}) |
397
|
|
|
|
|
|
|
or croak "No features to save"; |
398
|
|
|
|
|
|
|
|
399
|
0
|
0
|
|
|
|
|
open my($fh), "> $file" or croak "Can't create $file: $!"; |
400
|
0
|
|
|
|
|
|
my $h = $f->as_hash; |
401
|
0
|
|
|
|
|
|
print $fh "# Total: ", $f->length, "\n"; |
402
|
|
|
|
|
|
|
|
403
|
0
|
|
|
|
|
|
foreach my $k (sort {$h->{$b} <=> $h->{$a}} keys %$h) { |
|
0
|
|
|
|
|
|
|
404
|
0
|
|
|
|
|
|
print $fh "$k\t$h->{$k}\n"; |
405
|
|
|
|
|
|
|
} |
406
|
0
|
|
|
|
|
|
close $fh; |
407
|
|
|
|
|
|
|
} |
408
|
|
|
|
|
|
|
|
409
|
|
|
|
|
|
|
sub restore_features { |
410
|
0
|
|
|
0
|
1
|
|
my ($self, $file, $n) = @_; |
411
|
|
|
|
|
|
|
|
412
|
0
|
0
|
|
|
|
|
open my($fh), "< $file" or croak "Can't open $file: $!"; |
413
|
|
|
|
|
|
|
|
414
|
0
|
|
|
|
|
|
my %hash; |
415
|
0
|
|
|
|
|
|
while (<$fh>) { |
416
|
0
|
0
|
|
|
|
|
next if /^#/; |
417
|
0
|
0
|
|
|
|
|
/^(.*)\t([\d.]+)$/ or croak "Malformed line: $_"; |
418
|
0
|
|
|
|
|
|
$hash{$1} = $2; |
419
|
0
|
0
|
0
|
|
|
|
last if defined $n and $. >= $n; |
420
|
|
|
|
|
|
|
} |
421
|
0
|
|
|
|
|
|
my $features = $self->create_delayed_object('features', features => \%hash); |
422
|
|
|
|
|
|
|
|
423
|
0
|
|
|
|
|
|
$self->delayed_object_params('document', use_features => $features); |
424
|
0
|
|
|
|
|
|
$self->delayed_object_params('collection', use_features => $features); |
425
|
|
|
|
|
|
|
} |
426
|
|
|
|
|
|
|
|
427
|
|
|
|
|
|
|
1; |
428
|
|
|
|
|
|
|
|
429
|
|
|
|
|
|
|
__END__ |