|
1
|
|
|
<?php declare(strict_types=1); |
|
2
|
|
|
|
|
3
|
|
|
namespace Goose\Modules\Extractors; |
|
4
|
|
|
|
|
5
|
|
|
use Goose\Article; |
|
6
|
|
|
use Goose\Traits\{NodeCommonTrait, NodeGravityTrait, ArticleMutatorTrait}; |
|
7
|
|
|
use Goose\Modules\{AbstractModule, ModuleInterface}; |
|
8
|
|
|
use DOMWrap\Element; |
|
9
|
|
|
|
|
10
|
|
|
/** |
|
11
|
|
|
* Content Extractor |
|
12
|
|
|
* |
|
13
|
|
|
* @package Goose\Modules\Extractors |
|
14
|
|
|
* @license http://www.apache.org/licenses/LICENSE-2.0 Apache License 2.0 |
|
15
|
|
|
*/ |
|
16
|
|
|
class ContentExtractor extends AbstractModule implements ModuleInterface { |
|
17
|
|
|
use ArticleMutatorTrait, NodeGravityTrait, NodeCommonTrait; |
|
18
|
|
|
|
|
19
|
|
|
/** @inheritdoc */ |
|
20
|
|
|
public function run(Article $article): self { |
|
21
|
|
|
$this->article($article); |
|
22
|
|
|
|
|
23
|
|
|
$article->setTopNode($this->getTopNode()); |
|
24
|
|
|
|
|
25
|
|
|
return $this; |
|
26
|
|
|
} |
|
27
|
|
|
|
|
28
|
|
|
/** |
|
29
|
|
|
* @param Article $article |
|
30
|
|
|
* |
|
31
|
|
|
* @return array |
|
32
|
|
|
*/ |
|
33
|
|
|
private function getTopNodeCandidatesByContents(Article $article): array { |
|
34
|
|
|
$results = []; |
|
35
|
|
|
|
|
36
|
|
|
$nodes = $article->getDoc()->find('p, td, pre'); |
|
37
|
|
|
|
|
38
|
|
|
foreach ($nodes as $node) { |
|
39
|
|
|
$wordStats = $this->config()->getStopWords()->getStopwordCount($node->text()); |
|
40
|
|
|
$highLinkDensity = $this->isHighLinkDensity($node); |
|
41
|
|
|
|
|
42
|
|
|
if ($wordStats->getStopWordCount() > 2 && !$highLinkDensity) { |
|
43
|
|
|
$results[] = $node; |
|
44
|
|
|
} |
|
45
|
|
|
} |
|
46
|
|
|
|
|
47
|
|
|
return $results; |
|
48
|
|
|
} |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* @param Element $node |
|
52
|
|
|
* @param int $i |
|
53
|
|
|
* @param int $totalNodes |
|
54
|
|
|
* |
|
55
|
|
|
* @return float |
|
56
|
|
|
*/ |
|
57
|
|
|
private function getTopNodeCandidateScore(Element $node, int $i, int $totalNodes): float { |
|
58
|
|
|
$boostScore = (1.0 / ($i + 1)) * 50; |
|
59
|
|
|
$bottomNodesForNegativeScore = $totalNodes * 0.25; |
|
60
|
|
|
|
|
61
|
|
|
if ($totalNodes > 15) { |
|
62
|
|
|
if ($totalNodes - $i <= $bottomNodesForNegativeScore) { |
|
63
|
|
|
$booster = $bottomNodesForNegativeScore - ($totalNodes - $i); |
|
64
|
|
|
$boostScore = pow($booster, 2) * -1; |
|
65
|
|
|
$negscore = abs($boostScore); |
|
66
|
|
|
if ($negscore > 40) { |
|
67
|
|
|
$boostScore = 5; |
|
68
|
|
|
} |
|
69
|
|
|
} |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
|
|
$wordStats = $this->config()->getStopWords()->getStopwordCount($node->text()); |
|
73
|
|
|
$upscore = $wordStats->getStopWordCount() + $boostScore; |
|
74
|
|
|
|
|
75
|
|
|
return $upscore; |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
/** |
|
79
|
|
|
* @param array $nodes |
|
80
|
|
|
* |
|
81
|
|
|
* @return Element|null |
|
82
|
|
|
*/ |
|
83
|
|
|
private function getTopNodeByScore(array $nodes): ?Element { |
|
84
|
|
|
$topNode = null; |
|
85
|
|
|
$topNodeScore = 0; |
|
86
|
|
|
|
|
87
|
|
|
foreach ($nodes as $node) { |
|
88
|
|
|
$score = $this->getScore($node); |
|
89
|
|
|
|
|
90
|
|
|
if ($score > $topNodeScore) { |
|
91
|
|
|
$topNode = $node; |
|
92
|
|
|
$topNodeScore = $score; |
|
93
|
|
|
} |
|
94
|
|
|
|
|
95
|
|
|
if ($topNode === false) { |
|
96
|
|
|
$topNode = $node; |
|
97
|
|
|
} |
|
98
|
|
|
} |
|
99
|
|
|
|
|
100
|
|
|
if ($topNode && $this->getScore($topNode) < 20) { |
|
101
|
|
|
return null; |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
|
|
return $topNode; |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
/** |
|
108
|
|
|
* @param Element $node |
|
109
|
|
|
* @param float $upscore |
|
110
|
|
|
* |
|
111
|
|
|
* @return self |
|
112
|
|
|
*/ |
|
113
|
|
|
private function calculateBestNodeCandidateScores(Element $node, float $upscore): self { |
|
114
|
|
|
if ($node->parent() instanceof Element) { |
|
115
|
|
|
$this->updateScore($node->parent(), $upscore); |
|
116
|
|
|
$this->updateNodeCount($node->parent(), 1); |
|
117
|
|
|
|
|
118
|
|
|
if ($node->parent()->parent() instanceof Element) { |
|
119
|
|
|
$this->updateScore($node->parent()->parent(), $upscore / 2); |
|
120
|
|
|
$this->updateNodeCount($node->parent()->parent(), 1); |
|
121
|
|
|
} |
|
122
|
|
|
} |
|
123
|
|
|
|
|
124
|
|
|
return $this; |
|
125
|
|
|
} |
|
126
|
|
|
|
|
127
|
|
|
/** |
|
128
|
|
|
* @param Element $node |
|
129
|
|
|
* @param array $nodeCandidates |
|
130
|
|
|
* |
|
131
|
|
|
* @return array |
|
132
|
|
|
*/ |
|
133
|
|
|
private function updateBestNodeCandidates(Element $node, array $nodeCandidates): array { |
|
134
|
|
|
if (!in_array($node->parent(), $nodeCandidates, true)) { |
|
135
|
|
|
if ($node->parent() instanceof Element) { |
|
136
|
|
|
$nodeCandidates[] = $node->parent(); |
|
137
|
|
|
} |
|
138
|
|
|
} |
|
139
|
|
|
|
|
140
|
|
|
if ($node->parent() instanceof Element) { |
|
141
|
|
|
if (!in_array($node->parent()->parent(), $nodeCandidates, true)) { |
|
142
|
|
|
if ($node->parent()->parent() instanceof Element) { |
|
143
|
|
|
$nodeCandidates[] = $node->parent()->parent(); |
|
144
|
|
|
} |
|
145
|
|
|
} |
|
146
|
|
|
} |
|
147
|
|
|
|
|
148
|
|
|
return $nodeCandidates; |
|
149
|
|
|
} |
|
150
|
|
|
|
|
151
|
|
|
/** |
|
152
|
|
|
* We're going to start looking for where the clusters of paragraphs are. We'll score a cluster based on the number of stopwords |
|
153
|
|
|
* and the number of consecutive paragraphs together, which should form the cluster of text that this node is around |
|
154
|
|
|
* also store on how high up the paragraphs are, comments are usually at the bottom and should get a lower score |
|
155
|
|
|
* |
|
156
|
|
|
* @return Element|null |
|
157
|
|
|
*/ |
|
158
|
|
|
public function getTopNode(): ?Element { |
|
159
|
|
|
$nodes = $this->getTopNodeCandidatesByContents($this->article()); |
|
160
|
|
|
|
|
161
|
|
|
$nodeCandidates = []; |
|
162
|
|
|
|
|
163
|
|
|
$i = 0; |
|
164
|
|
|
foreach ($nodes as $node) { |
|
165
|
|
|
if ($this->isOkToBoost($node)) { |
|
166
|
|
|
$upscore = $this->getTopNodeCandidateScore($node, $i, count($nodes)); |
|
167
|
|
|
|
|
168
|
|
|
$this->calculateBestNodeCandidateScores($node, $upscore); |
|
169
|
|
|
$nodeCandidates = $this->updateBestNodeCandidates($node, $nodeCandidates); |
|
170
|
|
|
|
|
171
|
|
|
$i++; |
|
172
|
|
|
} |
|
173
|
|
|
} |
|
174
|
|
|
|
|
175
|
|
|
return $this->getTopNodeByScore($nodeCandidates); |
|
176
|
|
|
} |
|
177
|
|
|
|
|
178
|
|
|
/** |
|
179
|
|
|
* A lot of times the first paragraph might be the caption under an image so we'll want to make sure if we're going to |
|
180
|
|
|
* boost a parent node that it should be connected to other paragraphs, at least for the first n paragraphs |
|
181
|
|
|
* so we'll want to make sure that the next sibling is a paragraph and has at least some substantial weight to it |
|
182
|
|
|
* |
|
183
|
|
|
* @param Element $node |
|
184
|
|
|
* |
|
185
|
|
|
* @return bool |
|
186
|
|
|
*/ |
|
187
|
|
|
private function isOkToBoost(Element $node): bool { |
|
188
|
|
|
$stepsAway = 0; |
|
189
|
|
|
$minimumStopWordCount = 5; |
|
190
|
|
|
$maxStepsAwayFromNode = 3; |
|
191
|
|
|
|
|
192
|
|
|
// Find all previous sibling element nodes |
|
193
|
|
|
$siblings = $node->precedingAll(function($node) { |
|
194
|
|
|
return $node instanceof Element; |
|
195
|
|
|
}); |
|
196
|
|
|
|
|
197
|
|
|
foreach ($siblings as $sibling) { |
|
198
|
|
|
if ($sibling->is('p, strong')) { |
|
199
|
|
|
if ($stepsAway >= $maxStepsAwayFromNode) { |
|
200
|
|
|
return false; |
|
201
|
|
|
} |
|
202
|
|
|
|
|
203
|
|
|
$wordStats = $this->config()->getStopWords()->getStopwordCount($sibling->text()); |
|
204
|
|
|
|
|
205
|
|
|
if ($wordStats->getStopWordCount() > $minimumStopWordCount) { |
|
206
|
|
|
return true; |
|
207
|
|
|
} |
|
208
|
|
|
|
|
209
|
|
|
$stepsAway += 1; |
|
210
|
|
|
} |
|
211
|
|
|
} |
|
212
|
|
|
|
|
213
|
|
|
return false; |
|
214
|
|
|
} |
|
215
|
|
|
} |
|
216
|
|
|
|
scotteh /
php-goose
