biblicit 2.0.3 → 2.0.4
Sign up to get free protection for your applications and to get access to all the features.
- data/README.md +0 -2
- data/biblicit.gemspec +1 -1
- data/parscit/bin/citeExtract.pl +9 -161
- data/parscit/bin/sectExtract.pl +0 -14
- data/parscit/lib/ParsCit/Controller.pm +0 -59
- data/parscit/lib/ParsCit/PreProcess.pm +0 -4
- data/parscit/lib/ParsCit/Tr2crfpp.pm +1 -7
- metadata +4 -24
- data/parscit/bin/sectLabel/processOmniXML.pl +0 -1427
- data/parscit/bin/sectLabel/processOmniXML_new.pl +0 -1025
- data/parscit/bin/sectLabel/processOmniXMLv2.pl +0 -1529
- data/parscit/bin/sectLabel/processOmniXMLv3.pl +0 -964
- data/parscit/bin/sectLabel/simplifyOmniXML.pl +0 -382
- data/parscit/bin/xml2train.pl +0 -193
- data/parscit/lib/Omni/Config.pm +0 -93
- data/parscit/lib/Omni/Omnicell.pm +0 -263
- data/parscit/lib/Omni/Omnicol.pm +0 -292
- data/parscit/lib/Omni/Omnidd.pm +0 -328
- data/parscit/lib/Omni/Omnidoc.pm +0 -153
- data/parscit/lib/Omni/Omniframe.pm +0 -223
- data/parscit/lib/Omni/Omniline.pm +0 -423
- data/parscit/lib/Omni/Omnipage.pm +0 -282
- data/parscit/lib/Omni/Omnipara.pm +0 -232
- data/parscit/lib/Omni/Omnirun.pm +0 -303
- data/parscit/lib/Omni/Omnitable.pm +0 -336
- data/parscit/lib/Omni/Omniword.pm +0 -162
- data/parscit/lib/Omni/Traversal.pm +0 -313
- data/parscit/lib/SectLabel/AAMatching.pm +0 -1949
@@ -1,313 +0,0 @@
|
|
1
|
-
package Omni::Traversal;
|
2
|
-
|
3
|
-
# Configuration
|
4
|
-
use strict;
|
5
|
-
|
6
|
-
# Local libraries
|
7
|
-
use Omni::Config;
|
8
|
-
|
9
|
-
# Omnilib configuration: object name
|
10
|
-
my $obj_list = $Omni::Config::obj_list;
|
11
|
-
|
12
|
-
###
|
13
|
-
# 16 Feb 2010: Do Hoang Nhat Huy
|
14
|
-
# Omni lib is used to handle the Omnipage XML in a generic
|
15
|
-
# and 'elegent' way. The hierachical structure of the classes
|
16
|
-
# reflects the XML tree as follow
|
17
|
-
# Omnidoc
|
18
|
-
# |
|
19
|
-
# Omnipage (obviously, one document can have many pages)
|
20
|
-
# |__________________________________
|
21
|
-
# || ||
|
22
|
-
# Omnicol (columns in a page) Omnidd (image?)
|
23
|
-
# | |
|
24
|
-
# | Omnidd (nested)
|
25
|
-
# |_________________________________|
|
26
|
-
# | |
|
27
|
-
# Omnipara (a paragraph) Omnitable (a table)
|
28
|
-
# | |
|
29
|
-
# | Omnicell (a cell)
|
30
|
-
# | |
|
31
|
-
# | Omnipara (a paragraph)
|
32
|
-
# |_________________________________|
|
33
|
-
# |
|
34
|
-
# Omniline (a line)
|
35
|
-
# |
|
36
|
-
# Omnirun (a run, text of the same format)
|
37
|
-
# |
|
38
|
-
# Omniword (an individual word)
|
39
|
-
#
|
40
|
-
# This module provide a generic way to travel the whole tree
|
41
|
-
# to access each line of the document. Each line will have a
|
42
|
-
# group of identification (index) as follow
|
43
|
-
# page id;
|
44
|
-
# column id or dd id;
|
45
|
-
# column id or dd id;*
|
46
|
-
# table id;
|
47
|
-
# cell id;
|
48
|
-
# para id;
|
49
|
-
# line id;
|
50
|
-
# *: both dd and column can be nested inside each other
|
51
|
-
###
|
52
|
-
|
53
|
-
###
|
54
|
-
# Huydhn: collect lines whose addresses are selected
|
55
|
-
###
|
56
|
-
sub OmniCollector
|
57
|
-
{
|
58
|
-
my ($doc, $line_addrs, $need_obj) = @_;
|
59
|
-
|
60
|
-
# All the line
|
61
|
-
my @line_content = ();
|
62
|
-
|
63
|
-
# Check the validity
|
64
|
-
if (scalar(@{ $line_addrs }) == 0) { return (\@line_content); }
|
65
|
-
|
66
|
-
# Current position
|
67
|
-
my %current = ();
|
68
|
-
#
|
69
|
-
my $addr_index = 0;
|
70
|
-
|
71
|
-
# All pages in the document
|
72
|
-
my $pages = $doc->get_objs_ref();
|
73
|
-
|
74
|
-
# From page, To page
|
75
|
-
my $start_page = $line_addrs->[ 0 ]->{ 'L1' };
|
76
|
-
my $end_page = $line_addrs->[ -1 ]->{ 'L1' };
|
77
|
-
|
78
|
-
# Break condition: $line_pos is empty or all lines have been retrieved
|
79
|
-
my $break = 0;
|
80
|
-
|
81
|
-
# Tree traveling is 'not' fun. Seriously.
|
82
|
-
# This is like a dungeon seige.
|
83
|
-
for (my $x = $start_page; $x <= $end_page; $x++)
|
84
|
-
{
|
85
|
-
# Column or dd
|
86
|
-
my $level_2 = $pages->[ $x ]->get_objs_ref();
|
87
|
-
my $start_l2 = ($x == $line_addrs->[ 0 ]->{ 'L1' }) ?
|
88
|
-
$line_addrs->[ 0 ]->{ 'L2' } : 0;
|
89
|
-
my $end_l2 = ($x == $line_addrs->[ -1 ]->{ 'L1' }) ?
|
90
|
-
$line_addrs->[ -1 ]->{ 'L2' } : (scalar(@{ $level_2 }) - 1);
|
91
|
-
|
92
|
-
for (my $y = $start_l2; $y <= $end_l2; $y++)
|
93
|
-
{
|
94
|
-
# Table or paragraph
|
95
|
-
my $level_3 = $level_2->[ $y ]->get_objs_ref();
|
96
|
-
my $start_l3 = (($x == $line_addrs->[ 0 ]->{ 'L1' }) && ($y == $line_addrs->[ 0 ]->{ 'L2' })) ?
|
97
|
-
$line_addrs->[ 0 ]->{ 'L3' } : 0;
|
98
|
-
my $end_l3 = (($x == $line_addrs->[ -1 ]->{ 'L1' }) && ($y == $line_addrs->[ -1 ]->{ 'L2' })) ?
|
99
|
-
$line_addrs->[ -1 ]->{ 'L3' } : (scalar(@{ $level_3 }) - 1);
|
100
|
-
|
101
|
-
for (my $z = $start_l3; $z <= $end_l3; $z++)
|
102
|
-
{
|
103
|
-
# Is a paragraph
|
104
|
-
if ($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNIPARA' })
|
105
|
-
{
|
106
|
-
# Line or cell
|
107
|
-
my $level_4 = $level_3->[ $z ]->get_objs_ref();
|
108
|
-
my $start_l4 = (($x == $line_addrs->[ 0 ]->{ 'L1' }) && ($y == $line_addrs->[ 0 ]->{ 'L2' }) && ($z == $line_addrs->[ 0 ]->{ 'L3' })) ?
|
109
|
-
$line_addrs->[ 0 ]->{ 'L4' } : 0;
|
110
|
-
my $end_l4 = (($x == $line_addrs->[ -1 ]->{ 'L1' }) && ($y == $line_addrs->[ -1 ]->{ 'L2' }) && ($z == $line_addrs->[ -1 ]->{ 'L3' })) ?
|
111
|
-
$line_addrs->[ -1 ]->{ 'L4' } : (scalar(@{ $level_4 }) - 1);
|
112
|
-
|
113
|
-
# Lines
|
114
|
-
for (my $t = $start_l4; $t <= $end_l4; $t++)
|
115
|
-
{
|
116
|
-
# Only keep selected line
|
117
|
-
if (($x == $line_addrs->[ $addr_index ]{ 'L1' }) &&
|
118
|
-
($y == $line_addrs->[ $addr_index ]{ 'L2' }) &&
|
119
|
-
($z == $line_addrs->[ $addr_index ]{ 'L3' }) &&
|
120
|
-
($t == $line_addrs->[ $addr_index ]{ 'L4' }))
|
121
|
-
{
|
122
|
-
if ((! defined $need_obj) || ($need_obj == 0))
|
123
|
-
{
|
124
|
-
push @line_content, $level_4->[ $t ]->get_content();
|
125
|
-
}
|
126
|
-
else
|
127
|
-
{
|
128
|
-
push @line_content, $level_4->[ $t ];
|
129
|
-
}
|
130
|
-
|
131
|
-
# Next selected line
|
132
|
-
$addr_index++;
|
133
|
-
# Last one?
|
134
|
-
if ($addr_index == scalar(@{ $line_addrs }))
|
135
|
-
{
|
136
|
-
$break = 1;
|
137
|
-
last;
|
138
|
-
}
|
139
|
-
}
|
140
|
-
}
|
141
|
-
}
|
142
|
-
# Is a table
|
143
|
-
elsif (($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNITABLE' }) || ($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNIFRAME' }))
|
144
|
-
{
|
145
|
-
# TODO: this actually a trick to get it working for now.
|
146
|
-
# We care not about the cell inside the table but the content
|
147
|
-
# of the table only. So the table is consider a paragraph in
|
148
|
-
# which lines are its row
|
149
|
-
my @level_4 = split(/\n/, $level_3->[ $z ]->get_content());
|
150
|
-
|
151
|
-
for (my $t = 0; $t < scalar(@level_4); $t++)
|
152
|
-
{
|
153
|
-
# Current position
|
154
|
-
$current{ 'L4' } = $t;
|
155
|
-
|
156
|
-
# Only keep selected line
|
157
|
-
if (($x == $line_addrs->[ $addr_index ]{ 'L1' }) &&
|
158
|
-
($y == $line_addrs->[ $addr_index ]{ 'L2' }) &&
|
159
|
-
($z == $line_addrs->[ $addr_index ]{ 'L3' }) &&
|
160
|
-
($t == $line_addrs->[ $addr_index ]{ 'L4' }))
|
161
|
-
{
|
162
|
-
if ((! defined $need_obj) || ($need_obj == 0)) { push @line_content, $level_4[ $t ]; }
|
163
|
-
# Next selected line
|
164
|
-
$addr_index++;
|
165
|
-
# Last one?
|
166
|
-
if ($addr_index == scalar(@{ $line_addrs }))
|
167
|
-
{
|
168
|
-
$break = 1;
|
169
|
-
last;
|
170
|
-
}
|
171
|
-
}
|
172
|
-
}
|
173
|
-
}
|
174
|
-
|
175
|
-
# Break or not
|
176
|
-
if ($break == 1) { last; }
|
177
|
-
}
|
178
|
-
|
179
|
-
# Break or not
|
180
|
-
if ($break == 1) { last; }
|
181
|
-
}
|
182
|
-
|
183
|
-
# Break or not
|
184
|
-
if ($break == 1) { last; }
|
185
|
-
}
|
186
|
-
|
187
|
-
return (\@line_content);
|
188
|
-
}
|
189
|
-
|
190
|
-
###
|
191
|
-
# Huydhn: travel the Omnidoc at line level
|
192
|
-
###
|
193
|
-
sub OmniAirline
|
194
|
-
{
|
195
|
-
# Omnidoc object
|
196
|
-
# Starting position
|
197
|
-
# Ending position
|
198
|
-
# Both positons are hash with following members
|
199
|
-
# 'L1' : page
|
200
|
-
# 'L2' : collumn or dd or frame
|
201
|
-
# 'L3' : table or paragraph or frame
|
202
|
-
# 'L4' : line in paragraph or table or frame
|
203
|
-
my ($doc, $start, $end) = @_;
|
204
|
-
|
205
|
-
# All the line
|
206
|
-
my @line_pos = ();
|
207
|
-
my @line_content = ();
|
208
|
-
|
209
|
-
# Current position
|
210
|
-
my %current = ();
|
211
|
-
|
212
|
-
# All pages in the document
|
213
|
-
my $pages = $doc->get_objs_ref();
|
214
|
-
|
215
|
-
# From page, To page
|
216
|
-
my $start_page = (defined $start) ? $start->{ 'L1' } : 0;
|
217
|
-
my $end_page = (defined $end) ? $end->{ 'L1' } : (scalar(@{ $pages }) - 1);
|
218
|
-
|
219
|
-
# Tree traveling is 'not' fun. Seriously.
|
220
|
-
# This is like a dungeon seige.
|
221
|
-
for (my $x = $start_page; $x <= $end_page; $x++)
|
222
|
-
{
|
223
|
-
# Current position
|
224
|
-
$current{ 'L1' } = $x;
|
225
|
-
|
226
|
-
# Column or dd
|
227
|
-
my $level_2 = $pages->[ $x ]->get_objs_ref();
|
228
|
-
my $start_l2 = ((defined $start) && ($x == $start->{ 'L1' })) ?
|
229
|
-
$start->{ 'L2' } : 0;
|
230
|
-
my $end_l2 = ((defined $end) && ($x == $end->{ 'L1' })) ?
|
231
|
-
$end->{ 'L2' } : (scalar(@{ $level_2 }) - 1);
|
232
|
-
|
233
|
-
for (my $y = $start_l2; $y <= $end_l2; $y++)
|
234
|
-
{
|
235
|
-
# Current position
|
236
|
-
$current{ 'L2' } = $y;
|
237
|
-
|
238
|
-
# Table or paragraph
|
239
|
-
my $level_3 = $level_2->[ $y ]->get_objs_ref();
|
240
|
-
my $start_l3 = ((defined $start) && ($x == $start->{ 'L1' }) && ($y == $start->{ 'L2' })) ?
|
241
|
-
$start->{ 'L3' } : 0;
|
242
|
-
my $end_l3 = ((defined $end) &&($x == $end->{ 'L1' }) && ($y == $end->{ 'L2' })) ?
|
243
|
-
$end->{ 'L3' } : (scalar(@{ $level_3 }) - 1);
|
244
|
-
|
245
|
-
for (my $z = $start_l3; $z <= $end_l3; $z++)
|
246
|
-
{
|
247
|
-
# Current position
|
248
|
-
$current{ 'L3' } = $z;
|
249
|
-
|
250
|
-
# Is a paragraph
|
251
|
-
if ($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNIPARA' })
|
252
|
-
{
|
253
|
-
# Line or cell
|
254
|
-
my $level_4 = $level_3->[ $z ]->get_objs_ref();
|
255
|
-
my $start_l4 = ((defined $start) && ($x == $start->{ 'L1' }) && ($y == $start->{ 'L2' }) && ($z == $start->{ 'L3' })) ?
|
256
|
-
$start->{ 'L4' } : 0;
|
257
|
-
my $end_l4 = ((defined $end) && ($x == $end->{ 'L1' }) && ($y == $end->{ 'L2' }) && ($z == $end->{ 'L3' })) ?
|
258
|
-
$end->{ 'L4' } : (scalar(@{ $level_4 }) - 1);
|
259
|
-
|
260
|
-
# Lines
|
261
|
-
for (my $t = $start_l4; $t <= $end_l4; $t++)
|
262
|
-
{
|
263
|
-
# Current position
|
264
|
-
$current{ 'L4' } = $t;
|
265
|
-
|
266
|
-
# Only keep non-empty line
|
267
|
-
my $l = $level_4->[ $t ]->get_content();
|
268
|
-
$l =~ s/^\s+|\s+$//g;
|
269
|
-
|
270
|
-
if ($l ne "")
|
271
|
-
{
|
272
|
-
# Save the current position and the content of the current line
|
273
|
-
push @line_pos, { %current };
|
274
|
-
push @line_content, $level_4->[ $t ]->get_content();
|
275
|
-
}
|
276
|
-
}
|
277
|
-
}
|
278
|
-
# Is a table or frame
|
279
|
-
elsif (($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNITABLE' }) || ($level_3->[ $z ]->get_name() eq $obj_list->{ 'OMNIFRAME' }))
|
280
|
-
{
|
281
|
-
# TODO: this actually a trick to get it working for now.
|
282
|
-
# We care not about the cell inside the table but the content
|
283
|
-
# of the table only. So the table is consider a paragraph in
|
284
|
-
# which lines are its row
|
285
|
-
my @level_4 = split(/\n/, $level_3->[ $z ]->get_content());
|
286
|
-
|
287
|
-
for (my $t = 0; $t <= scalar(@level_4); $t++)
|
288
|
-
{
|
289
|
-
# Current position
|
290
|
-
$current{ 'L4' } = $t;
|
291
|
-
|
292
|
-
# Only keep non-empty line
|
293
|
-
my $l = $level_4[ $t ];
|
294
|
-
$l =~ s/^\s+|\s+$//g;
|
295
|
-
|
296
|
-
if ($l ne "")
|
297
|
-
{
|
298
|
-
# Save the current position and the content of the current line
|
299
|
-
push @line_pos, { %current };
|
300
|
-
push @line_content, $level_4[ $t ];
|
301
|
-
}
|
302
|
-
}
|
303
|
-
}
|
304
|
-
|
305
|
-
|
306
|
-
}
|
307
|
-
}
|
308
|
-
}
|
309
|
-
|
310
|
-
return (\@line_pos, \@line_content);
|
311
|
-
}
|
312
|
-
|
313
|
-
1;
|