|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Onoi\Tesa\Tokenizer; |
|
4
|
|
|
|
|
5
|
|
|
/** |
|
6
|
|
|
* @license GNU GPL v2+ |
|
7
|
|
|
* @since 0.1 |
|
8
|
|
|
* |
|
9
|
|
|
* @author mwjames |
|
10
|
|
|
*/ |
|
11
|
|
|
class NGramTokenizer implements Tokenizer { |
|
12
|
|
|
|
|
13
|
|
|
/** |
|
14
|
|
|
* @var Tokenizer |
|
15
|
|
|
*/ |
|
16
|
|
|
private $tokenizer; |
|
17
|
|
|
|
|
18
|
|
|
/** |
|
19
|
|
|
* @var integer |
|
20
|
|
|
*/ |
|
21
|
|
|
private $ngramSize = 2; |
|
22
|
|
|
|
|
23
|
|
|
/** |
|
24
|
|
|
* @var boolean |
|
25
|
|
|
*/ |
|
26
|
|
|
private $withMarker = false; |
|
27
|
|
|
|
|
28
|
|
|
/** |
|
29
|
|
|
* @since 0.1 |
|
30
|
|
|
* |
|
31
|
|
|
* @param Tokenizer $tokenizer |
|
32
|
|
|
* @param integer $ngramSize |
|
33
|
|
|
*/ |
|
34
|
|
|
public function __construct( Tokenizer $tokenizer = null, $ngramSize = 2 ) { |
|
35
|
|
|
$this->tokenizer = $tokenizer; |
|
36
|
|
|
$this->ngramSize = (int)$ngramSize; |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* @since 0.1 |
|
41
|
|
|
* |
|
42
|
|
|
* @param boolean $withMarker |
|
43
|
|
|
*/ |
|
44
|
|
|
public function withMarker( $withMarker ) { |
|
45
|
|
|
$this->withMarker = (bool)$withMarker; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
/** |
|
49
|
|
|
* @since 0.1 |
|
50
|
|
|
* |
|
51
|
|
|
* @param integer $ngramSize |
|
52
|
|
|
*/ |
|
53
|
|
|
public function setNgramSize( $ngramSize ) { |
|
54
|
|
|
$this->ngramSize = (int)$ngramSize; |
|
55
|
|
|
} |
|
56
|
|
|
|
|
57
|
|
|
/** |
|
58
|
|
|
* @since 0.1 |
|
59
|
|
|
* |
|
60
|
|
|
* {@inheritDoc} |
|
61
|
|
|
*/ |
|
62
|
|
|
public function setOption( $name, $value ) { |
|
63
|
|
|
if ( $this->tokenizer !== null ) { |
|
64
|
|
|
$this->tokenizer->setOption( $name, $value ); |
|
65
|
|
|
} |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
/** |
|
69
|
|
|
* @since 0.1 |
|
70
|
|
|
* |
|
71
|
|
|
* {@inheritDoc} |
|
72
|
|
|
*/ |
|
73
|
|
|
public function isWordTokenizer() { |
|
74
|
|
|
return false; |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
/** |
|
78
|
|
|
* @since 0.1 |
|
79
|
|
|
* |
|
80
|
|
|
* {@inheritDoc} |
|
81
|
|
|
*/ |
|
82
|
|
|
public function tokenize( $string ) { |
|
83
|
|
|
|
|
84
|
|
|
if ( $this->tokenizer !== null ) { |
|
85
|
|
|
$string = implode( " ", $this->tokenizer->tokenize( $string ) ); |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
$result = $this->createNGrams( $string, $this->ngramSize, $this->withMarker ); |
|
89
|
|
|
|
|
90
|
|
|
if ( $result !== false ) { |
|
91
|
|
|
return $result; |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
return array(); |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
private function createNGrams( $text, $ngramSize, $withMarker ) { |
|
98
|
|
|
|
|
99
|
|
|
$ngramList = []; |
|
100
|
|
|
|
|
101
|
|
|
// Identify the beginning-of-word and end-of-word |
|
102
|
|
|
if ( $withMarker ) { |
|
103
|
|
|
$text = '_' . str_replace( ' ', '_', $text ); |
|
104
|
|
|
} |
|
105
|
|
|
|
|
106
|
|
|
$text = mb_strtolower( $text ); |
|
107
|
|
|
$textLength = mb_strlen( $text, 'UTF-8' ); |
|
108
|
|
|
|
|
109
|
|
|
for ( $i = 0; $i < $textLength; ++$i ) { |
|
110
|
|
|
|
|
111
|
|
|
// Those failing the length of the ngramSize are skipped |
|
112
|
|
|
if ( !$withMarker && $i + $ngramSize > $textLength ) { |
|
113
|
|
|
continue; |
|
114
|
|
|
} |
|
115
|
|
|
|
|
116
|
|
|
$ngram = mb_substr( $text, $i, $ngramSize, 'UTF-8' ); |
|
117
|
|
|
|
|
118
|
|
|
// str_pad has issues with utf-8 length |
|
119
|
|
|
if ( $withMarker && ( $ngl = mb_strlen( $ngram, 'UTF-8' ) ) < $ngramSize ) { |
|
120
|
|
|
$ngram = $ngram . str_repeat( '_', ( $ngramSize - $ngl ) ); |
|
121
|
|
|
} |
|
122
|
|
|
|
|
123
|
|
|
$ngramList[] = $ngram; |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
return array_values( $ngramList ); |
|
127
|
|
|
} |
|
128
|
|
|
|
|
129
|
|
|
} |
|
130
|
|
|
|
onoi /
tesa
