|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Onoi\Tesa; |
|
4
|
|
|
|
|
5
|
|
|
use Onoi\Tesa\Tokenizer\Tokenizer; |
|
6
|
|
|
use Onoi\Tesa\Synonymizer\Synonymizer; |
|
7
|
|
|
use Onoi\Tesa\StopwordAnalyzer\StopwordAnalyzer; |
|
8
|
|
|
use RuntimeException; |
|
9
|
|
|
|
|
10
|
|
|
/** |
|
11
|
|
|
* @license GNU GPL v2+ |
|
12
|
|
|
* @since 0.1 |
|
13
|
|
|
* |
|
14
|
|
|
* @author mwjames |
|
15
|
|
|
*/ |
|
16
|
|
|
class Normalizer { |
|
17
|
|
|
|
|
18
|
|
|
/** |
|
19
|
|
|
* @since 0.1 |
|
20
|
|
|
* |
|
21
|
|
|
* @param string $text |
|
22
|
|
|
* |
|
23
|
|
|
* @param integer $flag |
|
24
|
|
|
*/ |
|
25
|
1 |
|
public static function applyTransliteration( $text, $flag = Transliterator::DIACRITICS ) { |
|
26
|
1 |
|
return Transliterator::transliterate( $text, $flag ); |
|
27
|
|
|
} |
|
28
|
|
|
|
|
29
|
|
|
/** |
|
30
|
|
|
* @see Localizer::convertDoubleWidth |
|
31
|
|
|
* |
|
32
|
|
|
* @since 0.1 |
|
33
|
|
|
* |
|
34
|
|
|
* @param string $text |
|
35
|
|
|
* |
|
36
|
|
|
* @return string |
|
37
|
|
|
*/ |
|
38
|
1 |
|
public static function convertDoubleWidth( $text ) { |
|
39
|
1 |
|
static $full = null; |
|
40
|
1 |
|
static $half = null; |
|
41
|
|
|
|
|
42
|
|
|
//,。／？《》〈〉；：“”＂〃＇｀［］｛｝＼｜～！－＝＿＋）（()＊…—─％￥＃ |
|
43
|
|
|
//,./?«»();:“” |
|
44
|
|
|
|
|
45
|
1 |
|
if ( $full === null ) { |
|
46
|
1 |
|
$fullWidth = "０１２３４５６７８９ＡＢＣＤＥＦＧＨＩＪＫＬＭＮＯＰＱＲＳＴＵＶＷＸＹＺａｂｃｄｅｆｇｈｉｊｋｌｍｎｏｐｑｒｓｔｕｖｗｘｙｚ"; |
|
47
|
1 |
|
$halfWidth = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"; |
|
48
|
|
|
|
|
49
|
|
|
// http://php.net/manual/en/function.str-split.php, mb_str_split |
|
50
|
1 |
|
$length = mb_strlen( $fullWidth, "UTF-8" ); |
|
51
|
1 |
|
$full = array(); |
|
52
|
|
|
|
|
53
|
1 |
|
for ( $i = 0; $i < $length; $i += 1 ) { |
|
54
|
1 |
|
$full[] = mb_substr( $fullWidth, $i, 1, "UTF-8" ); |
|
55
|
1 |
|
} |
|
56
|
|
|
|
|
57
|
1 |
|
$half = str_split( $halfWidth ); |
|
58
|
1 |
|
} |
|
59
|
|
|
|
|
60
|
1 |
|
return str_replace( $full, $half, trim( $text ) ); |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
|
|
64
|
|
|
/** |
|
65
|
|
|
* @since 0.1 |
|
66
|
|
|
* |
|
67
|
|
|
* @param string $text |
|
68
|
|
|
* |
|
69
|
|
|
* @return string |
|
70
|
|
|
*/ |
|
71
|
5 |
|
public static function toLowercase( $text ) { |
|
72
|
5 |
|
return mb_strtolower( $text, mb_detect_encoding( $text ) ); |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
|
|
/** |
|
76
|
|
|
* @since 0.1 |
|
77
|
|
|
* |
|
78
|
|
|
* @param string $text |
|
79
|
|
|
* @param integer|null $length |
|
80
|
|
|
* |
|
81
|
|
|
* @return string |
|
82
|
|
|
*/ |
|
83
|
1 |
|
public static function reduceLengthTo( $text, $length = null ) { |
|
84
|
|
|
|
|
85
|
1 |
|
if ( $length === null || mb_strlen( $text ) <= $length ) { |
|
86
|
1 |
|
return $text; |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
1 |
|
$encoding = mb_detect_encoding( $text ); |
|
90
|
1 |
|
$lastWholeWordPosition = $length; |
|
91
|
|
|
|
|
92
|
1 |
|
if ( strpos( $text, ' ' ) !== false ) { |
|
93
|
1 |
|
$lastWholeWordPosition = strrpos( mb_substr( $text, 0, $length, $encoding ), ' ' ); // last whole word |
|
94
|
1 |
|
} |
|
95
|
|
|
|
|
96
|
1 |
|
if ( $lastWholeWordPosition > 0 ) { |
|
97
|
1 |
|
$length = $lastWholeWordPosition; |
|
98
|
1 |
|
} |
|
99
|
|
|
|
|
100
|
1 |
|
return mb_substr( $text, 0, $length, $encoding ); |
|
101
|
|
|
} |
|
102
|
|
|
|
|
103
|
|
|
} |
|
104
|
|
|
|
onoi /
tesa
