|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
/** |
|
4
|
|
|
* |
|
5
|
|
|
* @Name : similar-text |
|
6
|
|
|
* @Programmer : Akpé Aurelle Emmanuel Moïse Zinsou |
|
7
|
|
|
* @Date : 2019-04-01 |
|
8
|
|
|
* @Released under : https://github.com/manuwhat/similar-text/blob/master/LICENSE |
|
9
|
|
|
* @Repository : https://github.com/manuwhat/similar |
|
10
|
|
|
* |
|
11
|
|
|
**/ |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
namespace EZAMA{ |
|
15
|
|
|
class distance extends complexCommonTextSimilarities |
|
16
|
|
|
{ |
|
17
|
|
|
public static function jaroWinkler($a, $b, $round=2) |
|
18
|
|
|
{ |
|
19
|
|
|
if (!is_string($a)||!is_string($b)) { |
|
20
|
|
|
return false; |
|
21
|
|
|
} |
|
22
|
|
|
static $distance=array(); |
|
23
|
|
|
static $previous=array(); |
|
24
|
|
|
if (array($a,$b)===$previous) { |
|
25
|
|
|
return $distance; |
|
26
|
|
|
} |
|
27
|
|
|
$previous=array($a,$b); |
|
28
|
|
|
return self::getJWDistance($a, $b, $distance, $round); |
|
29
|
|
|
} |
|
30
|
|
|
|
|
31
|
|
|
|
|
32
|
|
|
|
|
33
|
|
|
private static function getJWDistance(&$a, &$b, &$distance, $round) |
|
34
|
|
|
{ |
|
35
|
|
|
extract(self::prepareJaroWinkler($a, $b)); |
|
36
|
|
|
for ($i=0,$min=min(count($a), count($b)),$t=0;$i<$min;$i++) { |
|
37
|
|
|
if ($a[$i]!==$b[$i]) { |
|
38
|
|
|
$t++; |
|
39
|
|
|
} |
|
40
|
|
|
} |
|
41
|
|
|
$t/=2; |
|
42
|
|
|
$distance['jaro']=1/3*($corresponding/$ca+$corresponding/$cb+($corresponding-$t)/$corresponding); |
|
43
|
|
|
$distance['jaro-winkler']=$distance['jaro']+(min($longCommonSubstr, 4)*0.1*(1-$distance['jaro'])); |
|
44
|
|
|
$distance=array_map(function ($v) use ($round) { |
|
45
|
|
|
return round($v, $round); |
|
46
|
|
|
}, $distance); |
|
47
|
|
|
|
|
48
|
|
|
return $distance; |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
private static function prepareJaroWinkler(&$a, &$b) |
|
52
|
|
|
{ |
|
53
|
|
|
$a=self::split($a); |
|
54
|
|
|
$b=self::split($b); |
|
55
|
|
|
$transpositions=array('a'=>array(),'b'=>array(),'corresponding'=>0,'longCommonSubstr'=>0,'ca'=>count($a),'cb'=>count($b)); |
|
|
|
|
|
|
56
|
|
|
$Δ=max($transpositions['ca'], $transpositions['cb'])/2-1; |
|
57
|
|
|
self::jwMatches($a, $b, $transpositions, $Δ); |
|
58
|
|
|
ksort($transpositions['a']); |
|
59
|
|
|
ksort($transpositions['b']); |
|
60
|
|
|
$transpositions['a']=array_values($transpositions['a']); |
|
61
|
|
|
$transpositions['b']=array_values($transpositions['b']); |
|
62
|
|
|
return $transpositions; |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
|
|
private static function jwMatches(&$a, &$b, &$transpositions, $Δ) |
|
66
|
|
|
{ |
|
67
|
|
|
foreach ($a as $ind=>$chr) { |
|
68
|
|
|
foreach ($b as $index=>$char) { |
|
69
|
|
|
self::_jwMatches($chr, $char, $index, $ind, $transpositions, $Δ); |
|
70
|
|
|
} |
|
71
|
|
|
} |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
|
|
private static function _jwMatches($chr, $char, $index, $ind, &$transpositions, $Δ) |
|
75
|
|
|
{ |
|
76
|
|
|
if ($chr===$char&&(abs($index-$ind)<=$Δ)) { |
|
77
|
|
|
if ($ind!==$index) { |
|
78
|
|
|
$transpositions['a'][$ind]=$chr; |
|
79
|
|
|
$transpositions['b'][$index]=$char; |
|
80
|
|
|
} else { |
|
81
|
|
|
if ($ind-1<=$transpositions['longCommonSubstr']) { |
|
82
|
|
|
$transpositions['longCommonSubstr']++; |
|
83
|
|
|
} |
|
84
|
|
|
} |
|
85
|
|
|
$transpositions['corresponding']++; |
|
86
|
|
|
} |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
|
|
|
|
90
|
|
|
public static function hamming($a, $b) |
|
91
|
|
|
{ |
|
92
|
|
|
if (!is_string($a)||!is_string($b)||(strlen($a)!==strlen($b))) { |
|
93
|
|
|
return false; |
|
94
|
|
|
} |
|
95
|
|
|
static $distance=0; |
|
96
|
|
|
static $previous=array(); |
|
97
|
|
|
if (array($a,$b)===$previous) { |
|
98
|
|
|
return $distance; |
|
99
|
|
|
} |
|
100
|
|
|
$previous=array($a,$b); |
|
101
|
|
|
$a=self::split($a); |
|
102
|
|
|
$b=self::split($b); |
|
103
|
|
|
$distance=count(array_diff_assoc($a, $b)); |
|
|
|
|
|
|
104
|
|
|
return $distance; |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
public static function dice($a, $b, $round=2) |
|
108
|
|
|
{ |
|
109
|
|
|
if (!is_string($a)||!is_string($b)) { |
|
110
|
|
|
return false; |
|
111
|
|
|
} |
|
112
|
|
|
if (empty($a)||empty($b)) { |
|
113
|
|
|
return 0.0; |
|
114
|
|
|
} |
|
115
|
|
|
if ($a===$b) { |
|
116
|
|
|
return 1.0; |
|
117
|
|
|
} |
|
118
|
|
|
|
|
119
|
|
|
static $distance=0; |
|
120
|
|
|
static $previous=array(); |
|
121
|
|
|
if (array($a,$b)===$previous) { |
|
122
|
|
|
return $distance; |
|
123
|
|
|
} |
|
124
|
|
|
$previous=array($a,$b); |
|
125
|
|
|
$a=self::split($a, 2); |
|
126
|
|
|
$b=self::split($b, 2); |
|
127
|
|
|
return self::getDiceDistance($distance, $a, $b, $round); |
|
|
|
|
|
|
128
|
|
|
} |
|
129
|
|
|
|
|
130
|
|
|
private static function getDiceDistance(&$distance, &$a, &$b, $round) |
|
131
|
|
|
{ |
|
132
|
|
|
$ca=($caGrams=count($a))*2-self::getEndStrLen($a); |
|
133
|
|
|
$cb=($cbGrams=count($b))*2-self::getEndStrLen($b); |
|
134
|
|
|
$distance=round(2*count($caGrams>$cbGrams?array_intersect($a, $b):array_intersect($b, $a))/($ca+$cb), $round); |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
private static function getEndStrLen($a) |
|
138
|
|
|
{ |
|
139
|
|
|
if (function_exists('array_key_last')) { |
|
140
|
|
|
$end=array_key_last($a); |
|
141
|
|
|
$end=(isset($end[1]))?0:1; |
|
142
|
|
|
} else { |
|
143
|
|
|
$end=end($a); |
|
144
|
|
|
$end=(isset($end[1]))?0:1; |
|
145
|
|
|
reset($a); |
|
146
|
|
|
} |
|
147
|
|
|
return $end; |
|
148
|
|
|
} |
|
149
|
|
|
|
|
150
|
|
|
public static function levenshtein($a, $b) |
|
151
|
|
|
{ |
|
152
|
|
|
if (!is_string($a)||!is_string($b)) { |
|
153
|
|
|
return false; |
|
154
|
|
|
} |
|
155
|
|
|
|
|
156
|
|
|
|
|
157
|
|
|
static $distance=0; |
|
158
|
|
|
static $previous=array(); |
|
159
|
|
|
if (array($a,$b)===$previous) { |
|
160
|
|
|
return $distance; |
|
161
|
|
|
} |
|
162
|
|
|
$previous=array($a,$b); |
|
163
|
|
|
$a=self::split($a); |
|
164
|
|
|
$b=self::split($b); |
|
165
|
|
|
$ca = count($a); |
|
|
|
|
|
|
166
|
|
|
$cb = count($b); |
|
167
|
|
|
$dis = range(0, $cb); |
|
168
|
|
|
self::BuildLevenshteinCostMatrix($a, $b, $ca, $cb, $dis); |
|
169
|
|
|
|
|
170
|
|
|
return $distance=$dis[$cb]; |
|
171
|
|
|
} |
|
172
|
|
|
|
|
173
|
|
|
|
|
174
|
|
|
public static function levenshteinDamerau($a, $b) |
|
175
|
|
|
{ |
|
176
|
|
|
if (!is_string($a)||!is_string($b)) { |
|
177
|
|
|
return false; |
|
178
|
|
|
} |
|
179
|
|
|
|
|
180
|
|
|
static $distance=0; |
|
181
|
|
|
static $previous=array(); |
|
182
|
|
|
if (array($a,$b)===$previous) { |
|
183
|
|
|
return $distance; |
|
184
|
|
|
} |
|
185
|
|
|
$previous=array($a,$b); |
|
186
|
|
|
$a=self::split($a); |
|
187
|
|
|
$b=self::split($b); |
|
188
|
|
|
$ca = count($a); |
|
|
|
|
|
|
189
|
|
|
$cb = count($b); |
|
190
|
|
|
$dis = range(0, $cb); |
|
191
|
|
|
self::BuildLevenshteinCostMatrix($a, $b, $ca, $cb, $dis, true); |
|
192
|
|
|
|
|
193
|
|
|
return $distance=$dis[$cb]; |
|
194
|
|
|
} |
|
195
|
|
|
|
|
196
|
|
|
private static function BuildLevenshteinCostMatrix($a, $b, $ca, $cb, &$dis, $damerau=false) |
|
197
|
|
|
{ |
|
198
|
|
|
$dis_new=array(); |
|
199
|
|
|
for ($x=1;$x<=$ca;$x++) { |
|
200
|
|
|
$dis_new[0]=$x; |
|
201
|
|
|
for ($y=1;$y<=$cb;$y++) { |
|
202
|
|
|
self::costMatrix($a, $b, $dis_new, $dis, $damerau, $x, $y); |
|
203
|
|
|
} |
|
204
|
|
|
$dis = $dis_new; |
|
205
|
|
|
} |
|
206
|
|
|
} |
|
207
|
|
|
|
|
208
|
|
|
private static function costMatrix(&$a, &$b, &$dis_new, &$dis, $damerau, $x, $y) |
|
209
|
|
|
{ |
|
210
|
|
|
$c = ($a[$x-1] == $b[$y-1])?0:1; |
|
211
|
|
|
$dis_new[$y] = min($dis[$y]+1, $dis_new[$y-1]+1, $dis[$y-1]+$c); |
|
212
|
|
|
if ($damerau) { |
|
213
|
|
|
if ($x > 1 && $y > 1 && $a[$x-1] == $b[$y-2] && $a[$x-2] == $b[$y-1]) { |
|
214
|
|
|
$dis_new[$y]= min($dis_new[$y-1], $dis[$y-3] + $c) ; |
|
215
|
|
|
} |
|
216
|
|
|
} |
|
217
|
|
|
} |
|
218
|
|
|
} |
|
219
|
|
|
|
|
220
|
|
|
|
|
221
|
|
|
} |
|
222
|
|
|
|
manuwhat /
similar-text
Loading history...