|
1
|
|
|
<?php |
|
2
|
|
|
/* |
|
3
|
|
|
* This file is part of dispositif/wikibot application (@github) |
|
4
|
|
|
* 2019-2023 © Philippe M./Irønie <[email protected]> |
|
5
|
|
|
* For the full copyright and MIT license information, view the license file. |
|
6
|
|
|
*/ |
|
7
|
|
|
|
|
8
|
|
|
declare(strict_types=1); |
|
9
|
|
|
|
|
10
|
|
|
namespace App\Domain\Predict; |
|
11
|
|
|
|
|
12
|
|
|
use App\Domain\Utils\WikiTextUtil; |
|
13
|
|
|
|
|
14
|
|
|
/** |
|
15
|
|
|
* Prediction around many authors in same string. |
|
16
|
|
|
*/ |
|
17
|
|
|
class PredictAuthors |
|
18
|
|
|
{ |
|
19
|
|
|
private $typoPredict; |
|
20
|
|
|
|
|
21
|
|
|
private $authors = []; |
|
22
|
|
|
|
|
23
|
|
|
public function __construct() |
|
24
|
|
|
{ |
|
25
|
|
|
$this->typoPredict = new TypoTokenizer(); |
|
26
|
|
|
} |
|
27
|
|
|
|
|
28
|
|
|
/** |
|
29
|
|
|
* Explode authors from string based on typo pattern recognition. |
|
30
|
|
|
* See analysis_pattern_auteurs.php for stats and corpus generation. |
|
31
|
|
|
*/ |
|
32
|
|
|
public function predictAuthorNames(string $string): ?array |
|
33
|
|
|
{ |
|
34
|
|
|
$pattern = $this->typoPredict->typoPatternFromAuthor($string); |
|
35
|
|
|
|
|
36
|
|
|
$this->patternAndValueToAuthors($pattern); |
|
37
|
|
|
|
|
38
|
|
|
return $this->authors; |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
|
|
private function patternAndValueToAuthors(array $pattern): void |
|
42
|
|
|
{ |
|
43
|
|
|
$val = $pattern['value']; |
|
44
|
|
|
switch ($pattern['pattern']) { |
|
45
|
|
|
case 'FIRSTUPPER': |
|
46
|
|
|
case 'ALLUPPER': |
|
47
|
|
|
$this->authors = [0 => $val[0]]; |
|
48
|
|
|
|
|
49
|
|
|
break; |
|
50
|
|
|
|
|
51
|
|
|
case 'FIRSTUPPER FIRSTUPPER': // Laurent Croizier |
|
52
|
|
|
case 'MIXED FIRSTUPPER': // Jean-Paul Marchand |
|
53
|
|
|
case 'INITIAL FIRSTUPPER': // Christian Le Boutellier |
|
54
|
|
|
case 'FIRSTUPPER MIXED': |
|
55
|
|
|
case 'FIRSTUPPER FIRSTUPPER BIBABREV': |
|
56
|
|
|
case 'FIRSTUPPER FIRSTUPPER PUNCTUATION BIBABREV': |
|
57
|
|
|
$this->authors = [0 => $val[0].' '.$val[1]]; |
|
58
|
|
|
|
|
59
|
|
|
break; |
|
60
|
|
|
|
|
61
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER': |
|
62
|
|
|
case 'FIRSTUPPER INITIAL FIRSTUPPER': |
|
63
|
|
|
case 'INITIAL INITIAL FIRSTUPPER': |
|
64
|
|
|
case 'FIRSTUPPER ALLLOWER FIRSTUPPER': |
|
65
|
|
|
$this->authors = [0 => $val[0].' '.$val[1].' '.$val[2]]; |
|
66
|
|
|
|
|
67
|
|
|
break; |
|
68
|
|
|
|
|
69
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER FIRSTUPPER': |
|
70
|
|
|
// [[Jean Julien Michel Savary]] |
|
71
|
|
|
case 'FIRSTUPPER FIRSTUPPER ALLLOWER FIRSTUPPER': // Abbé Guillotin de Corson |
|
72
|
|
|
case 'FIRSTUPPER INITIAL INITIAL FIRSTUPPER': |
|
73
|
|
|
$this->authors = [0 => $val[0].' '.$val[1].' '.$val[2].' '.$val[3]]; |
|
74
|
|
|
|
|
75
|
|
|
break; |
|
76
|
|
|
|
|
77
|
|
|
// NOBLESSE |
|
78
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER ALLLOWER FIRSTUPPER': |
|
79
|
|
|
// [[Toussaint Du Breil de Pontbriand]] |
|
80
|
|
|
case 'MIXED FIRSTUPPER ALLLOWER FIRSTUPPER FIRSTUPPER': |
|
81
|
|
|
// [[Pierre-Suzanne Lucas de La Championnière]] |
|
82
|
|
|
case 'FIRSTUPPER ALLLOWER FIRSTUPPER ALLLOWER FIRSTUPPER': |
|
83
|
|
|
// [[Toussaint du Breil de Pontbriand]] |
|
84
|
|
|
case 'MIXED FIRSTUPPER FIRSTUPPER ALLLOWER FIRSTUPPER': |
|
85
|
|
|
// Marie-Paul Du Breil de Pontbriand |
|
86
|
|
|
case 'MIXED ALLLOWER FIRSTUPPER ALLLOWER FIRSTUPPER': |
|
87
|
|
|
// Marie-Paul du Breil de Pontbriand |
|
88
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER FIRSTUPPER FIRSTUPPER': |
|
89
|
|
|
// Mohamed El Aziz Ben Achour |
|
90
|
|
|
$this->authors = [0 => $val[0].' '.$val[1].' '.$val[2].' '.$val[3].' '.$val[4]]; |
|
91
|
|
|
|
|
92
|
|
|
break; |
|
93
|
|
|
|
|
94
|
|
|
/* |
|
95
|
|
|
* 2 authors |
|
96
|
|
|
*/ |
|
97
|
|
|
|
|
98
|
|
|
case 'FIRSTUPPER FIRSTUPPER AND FIRSTUPPER FIRSTUPPER': |
|
99
|
|
|
// Robert Sablayrolles et Argitxu Beyrie |
|
100
|
|
|
case 'MIXED FIRSTUPPER AND FIRSTUPPER FIRSTUPPER': |
|
101
|
|
|
case 'FIRSTUPPER FIRSTUPPER AND MIXED FIRSTUPPER': |
|
102
|
|
|
// Francine Musquère et Jean-Michel Mure |
|
103
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
104
|
|
|
// Annie Lagueyrie, Philippe Maviel |
|
105
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA MIXED FIRSTUPPER': |
|
106
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER MIXED': |
|
107
|
|
|
case 'MIXED FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
108
|
|
|
case 'INITIAL FIRSTUPPER COMMA INITIAL FIRSTUPPER': |
|
109
|
|
|
case 'FIRSTUPPER MIXED AND FIRSTUPPER FIRSTUPPER': |
|
110
|
|
|
$this->authors = [ |
|
111
|
|
|
0 => $val[0].' '.$val[1], |
|
112
|
|
|
1 => $val[3].' '.$val[4], |
|
113
|
|
|
]; |
|
114
|
|
|
|
|
115
|
|
|
break; |
|
116
|
|
|
|
|
117
|
|
|
// COUPLE |
|
118
|
|
|
case 'FIRSTUPPER AND FIRSTUPPER FIRSTUPPER': |
|
119
|
|
|
// Renée & Michel Paquet |
|
120
|
|
|
$this->authors = [ |
|
121
|
|
|
0 => $val[0].' '.$val[3], |
|
122
|
|
|
1 => $val[2].' '.$val[3], |
|
123
|
|
|
]; |
|
124
|
|
|
|
|
125
|
|
|
break; |
|
126
|
|
|
|
|
127
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
128
|
|
|
case 'FIRSTUPPER FIRSTUPPER FIRSTUPPER AND FIRSTUPPER FIRSTUPPER': |
|
129
|
|
|
// Didier Du Castel, Claude Estebe |
|
130
|
|
|
case 'FIRSTUPPER INITIAL FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
131
|
|
|
$this->authors = [ |
|
132
|
|
|
0 => $val[0].' '.$val[1].' '.$val[2], |
|
133
|
|
|
1 => $val[4].' '.$val[5], |
|
134
|
|
|
]; |
|
135
|
|
|
|
|
136
|
|
|
break; |
|
137
|
|
|
|
|
138
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER FIRSTUPPER': |
|
139
|
|
|
// Armin Vit, Bryony Gomez Palacio |
|
140
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER INITIAL FIRSTUPPER': |
|
141
|
|
|
case 'FIRSTUPPER FIRSTUPPER AND FIRSTUPPER FIRSTUPPER FIRSTUPPER': |
|
142
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER ALLLOWER MIXED': |
|
143
|
|
|
case 'FIRSTUPPER FIRSTUPPER AND FIRSTUPPER INITIAL FIRSTUPPER': |
|
144
|
|
|
// François Clément, Viton de Saint-Allais |
|
145
|
|
|
$this->authors = [ |
|
146
|
|
|
0 => $val[0].' '.$val[1], |
|
147
|
|
|
1 => $val[3].' '.$val[4].' '.$val[5], |
|
148
|
|
|
]; |
|
149
|
|
|
|
|
150
|
|
|
break; |
|
151
|
|
|
|
|
152
|
|
|
case 'FIRSTUPPER INITIAL FIRSTUPPER COMMA FIRSTUPPER INITIAL FIRSTUPPER': |
|
153
|
|
|
case 'INITIAL INITIAL FIRSTUPPER COMMA INITIAL INITIAL FIRSTUPPER': |
|
154
|
|
|
// Eugene P. Kiver, David V. Harris |
|
155
|
|
|
case 'INITIAL FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER INITIAL FIRSTUPPER': |
|
156
|
|
|
// H. Trevor Clifford, Peter D. Bostock |
|
157
|
|
|
case 'FIRSTUPPER INITIAL FIRSTUPPER AND FIRSTUPPER INITIAL FIRSTUPPER': |
|
158
|
|
|
case 'INITIAL INITIAL FIRSTUPPER AND INITIAL INITIAL FIRSTUPPER': |
|
159
|
|
|
$this->authors = [ |
|
160
|
|
|
0 => $val[0].' '.$val[1].' '.$val[2], |
|
161
|
|
|
1 => $val[4].' '.$val[5].' '.$val[6], |
|
162
|
|
|
]; |
|
163
|
|
|
|
|
164
|
|
|
break; |
|
165
|
|
|
|
|
166
|
|
|
/* |
|
167
|
|
|
* 3 authors |
|
168
|
|
|
*/ // Geddes, Czapor, Labahn |
|
169
|
|
|
case 'FIRSTUPPER COMMA FIRSTUPPER COMMA FIRSTUPPER': |
|
170
|
|
|
$this->authors = [ |
|
171
|
|
|
0 => $val[0], |
|
172
|
|
|
1 => $val[2], |
|
173
|
|
|
2 => $val[4], |
|
174
|
|
|
]; |
|
175
|
|
|
|
|
176
|
|
|
break; |
|
177
|
|
|
|
|
178
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER COMMA MIXED FIRSTUPPER': |
|
179
|
|
|
// Andrzej Suchcitz, Ludwik Maik, Wojciech Rojek |
|
180
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
181
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER AND FIRSTUPPER FIRSTUPPER': |
|
182
|
|
|
// [[Arnaud Bédat]], Gilles Bouleau et Bernard Nicolas |
|
183
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA MIXED FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
184
|
|
|
$this->authors = [ |
|
185
|
|
|
0 => $val[0].' '.$val[1], |
|
186
|
|
|
1 => $val[3].' '.$val[4], |
|
187
|
|
|
2 => $val[6].' '.$val[7], |
|
188
|
|
|
]; |
|
189
|
|
|
|
|
190
|
|
|
break; |
|
191
|
|
|
|
|
192
|
|
|
/* |
|
193
|
|
|
* 4 authors |
|
194
|
|
|
*/ |
|
195
|
|
|
|
|
196
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA MIXED FIRSTUPPER COMMA MIXED FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
197
|
|
|
case 'FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER COMMA FIRSTUPPER FIRSTUPPER': |
|
198
|
|
|
$this->authors = [ |
|
199
|
|
|
0 => $val[0].' '.$val[1], |
|
200
|
|
|
1 => $val[3].' '.$val[4], |
|
201
|
|
|
2 => $val[6].' '.$val[7], |
|
202
|
|
|
3 => $val[9].' '.$val[10], |
|
203
|
|
|
]; |
|
204
|
|
|
|
|
205
|
|
|
break; |
|
206
|
|
|
} |
|
207
|
|
|
} |
|
208
|
|
|
|
|
209
|
|
|
/** |
|
210
|
|
|
* From underTwoAuthors() by MartinS@Wikipedia |
|
211
|
|
|
* Return true if 0 or 1 author in $author; false otherwise. |
|
212
|
|
|
*/ |
|
213
|
|
|
public static function hasManyAuthors(string $author): bool |
|
214
|
|
|
{ |
|
215
|
|
|
$author = WikiTextUtil::unWikify($author); |
|
216
|
|
|
$chars = count_chars(trim($author)); |
|
217
|
|
|
|
|
218
|
|
|
// todo : "et" + "and" ? |
|
219
|
|
|
return $chars[ord('&')] > 0 || $chars[ord(';')] > 0 || $chars[ord(' ')] >= 3 || $chars[ord(',')] > 1; |
|
220
|
|
|
} |
|
221
|
|
|
} |
|
222
|
|
|
|