|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace rOpenDev\ExtractExpression; |
|
4
|
|
|
|
|
5
|
|
|
// curl: automatic set referrer |
|
6
|
|
|
// pomme de terre ne marche pas |
|
7
|
|
|
|
|
8
|
|
|
class ExtractExpressions |
|
9
|
|
|
{ |
|
10
|
|
|
public $onlyInSentence = false; |
|
11
|
|
|
public $expressionMaxWords = 5; |
|
12
|
|
|
public $keepTrail = 5; |
|
13
|
|
|
|
|
14
|
|
|
protected $text; |
|
15
|
|
|
|
|
16
|
|
|
protected $expressions = []; |
|
17
|
|
|
|
|
18
|
|
|
protected $wordNumber = 0; |
|
19
|
|
|
|
|
20
|
|
|
protected $trail = []; |
|
21
|
|
|
|
|
22
|
|
|
public function __construct(string $text) |
|
23
|
|
|
{ |
|
24
|
|
|
$this->addText($text); |
|
25
|
|
|
} |
|
26
|
|
|
|
|
27
|
|
|
protected function addText(string $text) |
|
28
|
|
|
{ |
|
29
|
|
|
$text = CleanText::stripHtmlTags($text); |
|
30
|
|
|
$text = CleanText::fixEncoding($text); |
|
31
|
|
|
|
|
32
|
|
|
$text = CleanText::removeDate($text); |
|
33
|
|
|
|
|
34
|
|
|
if ($this->onlyInSentence) { |
|
35
|
|
|
$text = CleanText::keepOnlySentence($text); |
|
36
|
|
|
} |
|
37
|
|
|
|
|
38
|
|
|
$this->text = $text; |
|
39
|
|
|
|
|
40
|
|
|
return $this; |
|
41
|
|
|
} |
|
42
|
|
|
|
|
43
|
|
|
public function getWordNumber() |
|
44
|
|
|
{ |
|
45
|
|
|
return $this->wordNumber; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
protected function incrementWordNumber($value) |
|
49
|
|
|
{ |
|
50
|
|
|
$this->wordNumber = $this->getWordNumber() + $value; |
|
51
|
|
|
} |
|
52
|
|
|
|
|
53
|
|
|
public function getExpressionsByDensity() |
|
54
|
|
|
{ |
|
55
|
|
|
$expressions = $this->expressions; |
|
56
|
|
|
|
|
57
|
|
|
foreach ($expressions as $k => $v) { |
|
58
|
|
|
$expressions[$k] = round(($v / $this->getWordNumber()) * 10000) / 100; |
|
59
|
|
|
} |
|
60
|
|
|
|
|
61
|
|
|
return $expressions; |
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
public function extract() |
|
65
|
|
|
{ |
|
66
|
|
|
if ($this->onlyInSentence) { |
|
67
|
|
|
$sentences = []; |
|
68
|
|
|
foreach (explode(chr(10), $this->text) as $paragraph) { |
|
69
|
|
|
$sentences = array_merge($sentences, CleanText::getSentences($paragraph)); |
|
70
|
|
|
} |
|
71
|
|
|
} else { |
|
72
|
|
|
$sentences = explode(chr(10), trim($this->text)); |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
|
|
foreach ($sentences as $sentence) { |
|
76
|
|
|
$sentence = CleanText::removePunctuation($sentence); |
|
77
|
|
|
|
|
78
|
|
|
$words = explode(' ', trim(strtolower($sentence))); |
|
79
|
|
|
|
|
80
|
|
|
foreach ($words as $key => $word) { |
|
81
|
|
|
for ($wordNumber = 1; $wordNumber < $this->expressionMaxWords; ++$wordNumber) { |
|
82
|
|
|
$expression = ''; |
|
83
|
|
|
for ($i = 0; $i < $wordNumber; ++$i) { |
|
84
|
|
|
if (isset($words[$key + $i])) { |
|
85
|
|
|
$expression .= ($i > 0 ? ' ' : '').$words[$key + $i]; |
|
86
|
|
|
} |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
|
|
$expression = $this->cleanExpr($expression, $wordNumber); |
|
90
|
|
|
|
|
91
|
|
|
if ( |
|
92
|
|
|
empty($expression) |
|
93
|
|
|
|| ((substr_count($expression, ' ') + 1) != $wordNumber) // We avoid sur-pondération |
|
94
|
|
|
|| !preg_match('/[a-z]/', $expression) // We avoid number or symbol only result |
|
95
|
|
|
) { |
|
96
|
|
|
if (1 === $wordNumber) { |
|
97
|
|
|
$this->incrementWordNumber(-1); |
|
98
|
|
|
} |
|
99
|
|
|
} else { |
|
100
|
|
|
$plus = 1 + substr_count(CleanText::removeStopWords($expression), ' '); |
|
101
|
|
|
$this->expressions[$expression] = isset($this->expressions[$expression]) ? $this->expressions[$expression] + $plus : $plus; |
|
102
|
|
|
if ($this->keepTrail > 0 && $this->expressions[$expression] > $this->keepTrail) { |
|
103
|
|
|
$this->trail[$expression][] = $sentence; |
|
104
|
|
|
} |
|
105
|
|
|
} |
|
106
|
|
|
} |
|
107
|
|
|
$this->incrementWordNumber(1); |
|
108
|
|
|
} |
|
109
|
|
|
} |
|
110
|
|
|
|
|
111
|
|
|
arsort($this->expressions); |
|
112
|
|
|
|
|
113
|
|
|
return $this->expressions; |
|
114
|
|
|
} |
|
115
|
|
|
|
|
116
|
|
|
protected function cleanExpr($expression, $wordNumber) |
|
117
|
|
|
{ |
|
118
|
|
|
if ($wordNumber <= 2) { |
|
119
|
|
|
$expression = trim(CleanText::removeStopWords(' '.$expression.' ')); |
|
120
|
|
|
} else { |
|
121
|
|
|
$expression = CleanText::removeStopWordsAtExtremity($expression); |
|
122
|
|
|
$expression = CleanText::removeStopWordsAtExtremity($expression); |
|
123
|
|
|
if (false === strpos($expression, ' ')) { |
|
124
|
|
|
$expression = trim(CleanText::removeStopWords(' '.$expression.' ')); |
|
125
|
|
|
} |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
// Last Clean |
|
129
|
|
|
$expression = trim(preg_replace('/\s+/', ' ', $expression)); |
|
130
|
|
|
if ('' == htmlentities($expression)) { //Avoid � |
|
131
|
|
|
$expression = ''; |
|
132
|
|
|
} |
|
133
|
|
|
|
|
134
|
|
|
return $expression; |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
public function getExpressions(?int $number = null) |
|
138
|
|
|
{ |
|
139
|
|
|
if (null === $this->expressions) { |
|
140
|
|
|
$this->extract(); |
|
141
|
|
|
} |
|
142
|
|
|
|
|
143
|
|
|
return !$number ? $this->expressions : array_slice($this->getExpressions(), 0, $number); |
|
144
|
|
|
} |
|
145
|
|
|
|
|
146
|
|
|
/** |
|
147
|
|
|
* @return array containing sentence where we can find expresion |
|
148
|
|
|
*/ |
|
149
|
|
|
public function getTrail(string $expression) |
|
150
|
|
|
{ |
|
151
|
|
|
if (null === $this->mergedExpressions) { |
|
|
|
|
|
|
152
|
|
|
$this->extract(); |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
if (isset($this->trail[$expression])) { |
|
156
|
|
|
return $this->trail[$expression]; |
|
157
|
|
|
} |
|
158
|
|
|
|
|
159
|
|
|
return []; |
|
160
|
|
|
} |
|
161
|
|
|
|
|
162
|
|
|
public function getTrails() |
|
163
|
|
|
{ |
|
164
|
|
|
return $this->trail; |
|
165
|
|
|
} |
|
166
|
|
|
} |
|
167
|
|
|
|
RobinDev /
ExtractExpression
Loading history...