|
1
|
|
|
<?php declare(strict_types=1); |
|
2
|
|
|
|
|
3
|
|
|
namespace Goose\Text; |
|
4
|
|
|
|
|
5
|
|
|
use Goose\Configuration; |
|
6
|
|
|
|
|
7
|
|
|
/** |
|
8
|
|
|
* Stop Words |
|
9
|
|
|
* |
|
10
|
|
|
* @package Goose\Text |
|
11
|
|
|
* @license http://www.apache.org/licenses/LICENSE-2.0 Apache License 2.0 |
|
12
|
|
|
*/ |
|
13
|
|
|
class StopWords |
|
14
|
|
|
{ |
|
15
|
|
|
/** @var Configuration */ |
|
16
|
|
|
private $config; |
|
17
|
|
|
|
|
18
|
|
|
/** @var string[] */ |
|
19
|
|
|
private $cached = []; |
|
20
|
|
|
|
|
21
|
|
|
/** @var string[] */ |
|
22
|
|
|
private $languages = [ |
|
23
|
|
|
'ar', 'da', 'de', 'en', 'es', 'fi', |
|
24
|
|
|
'fr', 'hu', 'id', 'it', 'ja', 'ko', |
|
25
|
|
|
'nb', 'nl', 'no', 'pl', 'pt', 'ru', |
|
26
|
|
|
'sv', 'zh' |
|
27
|
|
|
]; |
|
28
|
|
|
|
|
29
|
|
|
/** |
|
30
|
|
|
* @param Configuration $config |
|
31
|
|
|
*/ |
|
32
|
|
|
public function __construct(Configuration $config) { |
|
33
|
|
|
$this->config = $config; |
|
34
|
|
|
} |
|
35
|
|
|
|
|
36
|
|
|
/** |
|
37
|
|
|
* @return Configuration |
|
38
|
|
|
*/ |
|
39
|
|
|
public function config(): Configuration { |
|
40
|
|
|
return $this->config; |
|
41
|
|
|
} |
|
42
|
|
|
|
|
43
|
|
|
/** |
|
44
|
|
|
* @param string $str |
|
45
|
|
|
* |
|
46
|
|
|
* @return string |
|
47
|
|
|
*/ |
|
48
|
|
|
public function removePunctuation(string $str): string { |
|
49
|
|
|
return preg_replace("/[[:punct:]]+/", '', $str); |
|
50
|
|
|
} |
|
51
|
|
|
|
|
52
|
|
|
/** |
|
53
|
|
|
* @return string |
|
54
|
|
|
*/ |
|
55
|
|
|
public function getLanguage(): string { |
|
56
|
|
|
list($language) = explode('-', $this->config()->get('language')); |
|
57
|
|
|
|
|
58
|
|
|
if (!in_array($language, $this->languages)) { |
|
59
|
|
|
$language = 'en'; |
|
60
|
|
|
} |
|
61
|
|
|
return mb_strtolower($language); |
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
/** |
|
65
|
|
|
* @return mixed |
|
66
|
|
|
*/ |
|
67
|
|
|
public function getWordList(): array { |
|
68
|
|
|
if (empty($this->cached)) { |
|
69
|
|
|
$file = sprintf(__DIR__ . '/../../resources/text/stopwords-%s.txt', $this->getLanguage()); |
|
70
|
|
|
|
|
71
|
|
|
$this->cached = explode("\n", str_replace(["\r\n", "\r"], "\n", file_get_contents($file))); |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
|
|
return $this->cached; |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
/** |
|
78
|
|
|
* @param string $content |
|
79
|
|
|
* |
|
80
|
|
|
* @return WordStats |
|
81
|
|
|
*/ |
|
82
|
|
|
public function getStopwordCount(string $content): WordStats { |
|
83
|
|
|
if (empty($content)) { |
|
84
|
|
|
return new WordStats(); |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
$strippedInput = $this->removePunctuation($content); |
|
88
|
|
|
$candidateWords = $this->getCandidateWords($strippedInput); |
|
89
|
|
|
|
|
90
|
|
|
$overlappingStopWords = []; |
|
91
|
|
|
foreach ($candidateWords as $w) { |
|
|
|
|
|
|
92
|
|
|
if (in_array(mb_strtolower($w), $this->getWordList())) { |
|
93
|
|
|
$overlappingStopWords[] = mb_strtolower($w); |
|
94
|
|
|
} |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
return new WordStats([ |
|
98
|
|
|
'wordCount' => count($candidateWords), |
|
99
|
|
|
'stopWordCount' => count($overlappingStopWords), |
|
100
|
|
|
'stopWords' => $overlappingStopWords, |
|
101
|
|
|
]); |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
|
|
/** |
|
105
|
|
|
* @param string $strippedInput |
|
106
|
|
|
* |
|
107
|
|
|
* @return array |
|
108
|
|
|
*/ |
|
109
|
|
|
public function getCandidateWords(string $strippedInput): string { |
|
110
|
|
|
// Simple separating words in Japanese. |
|
111
|
|
|
if ($this->getLanguage() === 'ja') { |
|
112
|
|
|
$regexp = '/(' . implode('|', array_map('preg_quote', $this->getWordList())) . ')/'; |
|
113
|
|
|
$strippedInput = preg_replace($regexp, ' $1 ', $strippedInput); |
|
114
|
|
|
} |
|
115
|
|
|
|
|
116
|
|
|
return explode(' ', $strippedInput); |
|
117
|
|
|
} |
|
118
|
|
|
} |
scotteh /
php-goose
Loading history...