|
1
|
|
|
# -*- coding: utf-8 -*- |
|
2
|
|
|
|
|
3
|
|
|
# Copyright 2014-2018 by Christopher C. Little. |
|
4
|
|
|
# This file is part of Abydos. |
|
5
|
|
|
# |
|
6
|
|
|
# Abydos is free software: you can redistribute it and/or modify |
|
7
|
|
|
# it under the terms of the GNU General Public License as published by |
|
8
|
|
|
# the Free Software Foundation, either version 3 of the License, or |
|
9
|
|
|
# (at your option) any later version. |
|
10
|
|
|
# |
|
11
|
|
|
# Abydos is distributed in the hope that it will be useful, |
|
12
|
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
13
|
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
14
|
|
|
# GNU General Public License for more details. |
|
15
|
|
|
# |
|
16
|
|
|
# You should have received a copy of the GNU General Public License |
|
17
|
|
|
# along with Abydos. If not, see <http://www.gnu.org/licenses/>. |
|
18
|
|
|
|
|
19
|
1 |
|
"""abydos.distance._jaro_winkler. |
|
20
|
|
|
|
|
21
|
|
|
The distance._JaroWinkler module implements distance metrics based on |
|
22
|
|
|
:cite:`Jaro:1989` and subsequent works: |
|
23
|
|
|
|
|
24
|
|
|
- Jaro distance |
|
25
|
|
|
- Jaro-Winkler distance |
|
26
|
|
|
""" |
|
27
|
|
|
|
|
28
|
1 |
|
from __future__ import ( |
|
29
|
|
|
absolute_import, |
|
30
|
|
|
division, |
|
31
|
|
|
print_function, |
|
32
|
|
|
unicode_literals, |
|
33
|
|
|
) |
|
34
|
|
|
|
|
35
|
1 |
|
from six.moves import range |
|
36
|
|
|
|
|
37
|
1 |
|
from ._distance import _Distance |
|
38
|
1 |
|
from ..tokenizer import QGrams |
|
39
|
|
|
|
|
40
|
1 |
|
__all__ = ['JaroWinkler', 'dist_jaro_winkler', 'sim_jaro_winkler'] |
|
41
|
|
|
|
|
42
|
|
|
|
|
43
|
1 |
|
class JaroWinkler(_Distance): |
|
|
|
|
|
|
44
|
|
|
"""Jaro-Winkler distance. |
|
45
|
|
|
|
|
46
|
|
|
Jaro(-Winkler) distance is a string edit distance initially proposed by |
|
47
|
|
|
Jaro and extended by Winkler :cite:`Jaro:1989,Winkler:1990`. |
|
48
|
|
|
|
|
49
|
|
|
This is Python based on the C code for strcmp95: |
|
50
|
|
|
http://web.archive.org/web/20110629121242/http://www.census.gov/geo/msb/stand/strcmp.c |
|
51
|
|
|
:cite:`Winkler:1994`. The above file is a US Government publication and, |
|
52
|
|
|
accordingly, in the public domain. |
|
53
|
|
|
""" |
|
54
|
|
|
|
|
55
|
1 |
|
def sim( |
|
|
|
|
|
|
56
|
|
|
self, |
|
|
|
|
|
|
57
|
|
|
src, |
|
|
|
|
|
|
58
|
|
|
tar, |
|
|
|
|
|
|
59
|
|
|
qval=1, |
|
|
|
|
|
|
60
|
|
|
mode='winkler', |
|
|
|
|
|
|
61
|
|
|
long_strings=False, |
|
|
|
|
|
|
62
|
|
|
boost_threshold=0.7, |
|
|
|
|
|
|
63
|
|
|
scaling_factor=0.1, |
|
|
|
|
|
|
64
|
|
|
): |
|
65
|
|
|
"""Return the Jaro or Jaro-Winkler similarity of two strings. |
|
66
|
|
|
|
|
67
|
|
|
Args: |
|
68
|
|
|
src (str): Source string for comparison |
|
69
|
|
|
tar (str): Target string for comparison |
|
70
|
|
|
qval (int): The length of each q-gram (defaults to 1: |
|
71
|
|
|
character-wise matching) |
|
72
|
|
|
mode (str): Indicates which variant of this distance metric to |
|
73
|
|
|
compute: |
|
74
|
|
|
- ``winkler`` -- computes the Jaro-Winkler distance |
|
75
|
|
|
(default) which increases the score for matches near the |
|
76
|
|
|
start of the word |
|
77
|
|
|
- ``jaro`` -- computes the Jaro distance |
|
78
|
|
|
long_strings (bool): Set to True to "Increase the probability of a |
|
79
|
|
|
match when the number of matched characters is large. This |
|
80
|
|
|
option allows for a little more tolerance when the strings are |
|
81
|
|
|
large. It is not an appropriate test when comparing fixed |
|
82
|
|
|
length fields such as phone and social security numbers." |
|
83
|
|
|
(Used in 'winkler' mode only.) |
|
84
|
|
|
boost_threshold (float): A value between 0 and 1, below which the |
|
85
|
|
|
Winkler boost is not applied (defaults to 0.7). (Used in |
|
86
|
|
|
'winkler' mode only.) |
|
87
|
|
|
scaling_factor (float): A value between 0 and 0.25, indicating by |
|
88
|
|
|
how much to boost scores for matching prefixes (defaults to |
|
89
|
|
|
0.1). (Used in 'winkler' mode only.) |
|
90
|
|
|
|
|
91
|
|
|
Returns: |
|
92
|
|
|
float: Jaro or Jaro-Winkler similarity |
|
93
|
|
|
|
|
94
|
|
|
Raises: |
|
95
|
|
|
ValueError: Unsupported boost_threshold assignment; boost_threshold |
|
96
|
|
|
must be between 0 and 1. |
|
97
|
|
|
ValueError: Unsupported scaling_factor assignment; scaling_factor |
|
98
|
|
|
must be between 0 and 0.25.' |
|
99
|
|
|
|
|
100
|
|
|
Examples: |
|
101
|
|
|
>>> round(sim_jaro_winkler('cat', 'hat'), 12) |
|
102
|
|
|
0.777777777778 |
|
103
|
|
|
>>> round(sim_jaro_winkler('Niall', 'Neil'), 12) |
|
104
|
|
|
0.805 |
|
105
|
|
|
>>> round(sim_jaro_winkler('aluminum', 'Catalan'), 12) |
|
106
|
|
|
0.60119047619 |
|
107
|
|
|
>>> round(sim_jaro_winkler('ATCG', 'TAGC'), 12) |
|
108
|
|
|
0.833333333333 |
|
109
|
|
|
|
|
110
|
|
|
>>> round(sim_jaro_winkler('cat', 'hat', mode='jaro'), 12) |
|
111
|
|
|
0.777777777778 |
|
112
|
|
|
>>> round(sim_jaro_winkler('Niall', 'Neil', mode='jaro'), 12) |
|
113
|
|
|
0.783333333333 |
|
114
|
|
|
>>> round(sim_jaro_winkler('aluminum', 'Catalan', mode='jaro'), 12) |
|
115
|
|
|
0.60119047619 |
|
116
|
|
|
>>> round(sim_jaro_winkler('ATCG', 'TAGC', mode='jaro'), 12) |
|
117
|
|
|
0.833333333333 |
|
118
|
|
|
|
|
119
|
|
|
""" |
|
120
|
1 |
|
if mode == 'winkler': |
|
121
|
1 |
|
if boost_threshold > 1 or boost_threshold < 0: |
|
122
|
1 |
|
raise ValueError( |
|
123
|
|
|
'Unsupported boost_threshold assignment; ' |
|
124
|
|
|
+ 'boost_threshold must be between 0 and 1.' |
|
125
|
|
|
) |
|
126
|
1 |
|
if scaling_factor > 0.25 or scaling_factor < 0: |
|
127
|
1 |
|
raise ValueError( |
|
128
|
|
|
'Unsupported scaling_factor assignment; ' |
|
129
|
|
|
+ 'scaling_factor must be between 0 and 0.25.' |
|
130
|
|
|
) |
|
131
|
|
|
|
|
132
|
1 |
|
if src == tar: |
|
133
|
1 |
|
return 1.0 |
|
134
|
|
|
|
|
135
|
1 |
|
src = QGrams(src.strip(), qval).ordered_list |
|
136
|
1 |
|
tar = QGrams(tar.strip(), qval).ordered_list |
|
137
|
|
|
|
|
138
|
1 |
|
lens = len(src) |
|
139
|
1 |
|
lent = len(tar) |
|
140
|
|
|
|
|
141
|
|
|
# If either string is blank - return - added in Version 2 |
|
142
|
1 |
|
if lens == 0 or lent == 0: |
|
143
|
1 |
|
return 0.0 |
|
144
|
|
|
|
|
145
|
1 |
|
if lens > lent: |
|
146
|
1 |
|
search_range = lens |
|
147
|
1 |
|
minv = lent |
|
148
|
|
|
else: |
|
149
|
1 |
|
search_range = lent |
|
150
|
1 |
|
minv = lens |
|
151
|
|
|
|
|
152
|
|
|
# Zero out the flags |
|
153
|
1 |
|
src_flag = [0] * search_range |
|
154
|
1 |
|
tar_flag = [0] * search_range |
|
155
|
1 |
|
search_range = max(0, search_range // 2 - 1) |
|
156
|
|
|
|
|
157
|
|
|
# Looking only within the search range, |
|
158
|
|
|
# count and flag the matched pairs. |
|
159
|
1 |
|
num_com = 0 |
|
160
|
1 |
|
yl1 = lent - 1 |
|
161
|
1 |
|
for i in range(lens): |
|
162
|
1 |
|
low_lim = (i - search_range) if (i >= search_range) else 0 |
|
163
|
1 |
|
hi_lim = (i + search_range) if ((i + search_range) <= yl1) else yl1 |
|
164
|
1 |
|
for j in range(low_lim, hi_lim + 1): |
|
165
|
1 |
|
if (tar_flag[j] == 0) and (tar[j] == src[i]): |
|
166
|
1 |
|
tar_flag[j] = 1 |
|
167
|
1 |
|
src_flag[i] = 1 |
|
168
|
1 |
|
num_com += 1 |
|
169
|
1 |
|
break |
|
170
|
|
|
|
|
171
|
|
|
# If no characters in common - return |
|
172
|
1 |
|
if num_com == 0: |
|
173
|
1 |
|
return 0.0 |
|
174
|
|
|
|
|
175
|
|
|
# Count the number of transpositions |
|
176
|
1 |
|
k = n_trans = 0 |
|
177
|
1 |
|
for i in range(lens): |
|
178
|
1 |
|
if src_flag[i] != 0: |
|
179
|
1 |
|
j = 0 |
|
180
|
1 |
|
for j in range(k, lent): # pragma: no branch |
|
181
|
1 |
|
if tar_flag[j] != 0: |
|
182
|
1 |
|
k = j + 1 |
|
183
|
1 |
|
break |
|
184
|
1 |
|
if src[i] != tar[j]: |
|
185
|
1 |
|
n_trans += 1 |
|
186
|
1 |
|
n_trans //= 2 |
|
187
|
|
|
|
|
188
|
|
|
# Main weight computation for Jaro distance |
|
189
|
1 |
|
weight = ( |
|
190
|
|
|
num_com / lens + num_com / lent + (num_com - n_trans) / num_com |
|
191
|
|
|
) |
|
192
|
1 |
|
weight /= 3.0 |
|
193
|
|
|
|
|
194
|
|
|
# Continue to boost the weight if the strings are similar |
|
195
|
|
|
# This is the Winkler portion of Jaro-Winkler distance |
|
196
|
1 |
|
if mode == 'winkler' and weight > boost_threshold: |
|
197
|
|
|
|
|
198
|
|
|
# Adjust for having up to the first 4 characters in common |
|
199
|
1 |
|
j = 4 if (minv >= 4) else minv |
|
200
|
1 |
|
i = 0 |
|
201
|
1 |
|
while (i < j) and (src[i] == tar[i]): |
|
202
|
1 |
|
i += 1 |
|
203
|
1 |
|
weight += i * scaling_factor * (1.0 - weight) |
|
204
|
|
|
|
|
205
|
|
|
# Optionally adjust for long strings. |
|
206
|
|
|
|
|
207
|
|
|
# After agreeing beginning chars, at least two more must agree and |
|
208
|
|
|
# the agreeing characters must be > .5 of remaining characters. |
|
209
|
1 |
|
if ( |
|
210
|
|
|
long_strings |
|
|
|
|
|
|
211
|
|
|
and (minv > 4) |
|
|
|
|
|
|
212
|
|
|
and (num_com > i + 1) |
|
|
|
|
|
|
213
|
|
|
and (2 * num_com >= minv + i) |
|
|
|
|
|
|
214
|
|
|
): |
|
215
|
1 |
|
weight += (1.0 - weight) * ( |
|
216
|
|
|
(num_com - i - 1) / (lens + lent - i * 2 + 2) |
|
217
|
|
|
) |
|
218
|
|
|
|
|
219
|
1 |
|
return weight |
|
220
|
|
|
|
|
221
|
|
|
|
|
222
|
1 |
|
def sim_jaro_winkler( |
|
|
|
|
|
|
223
|
|
|
src, |
|
|
|
|
|
|
224
|
|
|
tar, |
|
|
|
|
|
|
225
|
|
|
qval=1, |
|
|
|
|
|
|
226
|
|
|
mode='winkler', |
|
|
|
|
|
|
227
|
|
|
long_strings=False, |
|
|
|
|
|
|
228
|
|
|
boost_threshold=0.7, |
|
|
|
|
|
|
229
|
|
|
scaling_factor=0.1, |
|
|
|
|
|
|
230
|
|
|
): |
|
231
|
|
|
"""Return the Jaro or Jaro-Winkler similarity of two strings. |
|
232
|
|
|
|
|
233
|
|
|
This is a wrapper for :py:meth:`JaroWinkler.sim`. |
|
234
|
|
|
|
|
235
|
|
|
Args: |
|
236
|
|
|
src (str): Source string for comparison |
|
237
|
|
|
tar (str): Target string for comparison |
|
238
|
|
|
qval (int): The length of each q-gram (defaults to 1: |
|
239
|
|
|
character-wise matching) |
|
240
|
|
|
mode (str): Indicates which variant of this distance metric to |
|
241
|
|
|
compute: |
|
242
|
|
|
- ``winkler`` -- computes the Jaro-Winkler distance (default) |
|
243
|
|
|
which increases the score for matches near the start of the |
|
244
|
|
|
word |
|
245
|
|
|
- ``jaro`` -- computes the Jaro distance |
|
246
|
|
|
long_strings (bool): Set to True to "Increase the probability of a |
|
247
|
|
|
match when the number of matched characters is large. This option |
|
248
|
|
|
allows for a little more tolerance when the strings are large. It |
|
249
|
|
|
is not an appropriate test when comparing fixedlength fields such |
|
250
|
|
|
as phone and social security numbers." (Used in 'winkler' mode |
|
251
|
|
|
only.) |
|
252
|
|
|
boost_threshold (float): A value between 0 and 1, below which the |
|
253
|
|
|
Winkler boost is not applied (defaults to 0.7). (Used in 'winkler' |
|
254
|
|
|
mode only.) |
|
255
|
|
|
scaling_factor (float): A value between 0 and 0.25, indicating by how |
|
256
|
|
|
much to boost scores for matching prefixes (defaults to 0.1). (Used |
|
257
|
|
|
in 'winkler' mode only.) |
|
258
|
|
|
|
|
259
|
|
|
Returns: |
|
260
|
|
|
float: Jaro or Jaro-Winkler similarity |
|
261
|
|
|
|
|
262
|
|
|
Examples: |
|
263
|
|
|
>>> round(sim_jaro_winkler('cat', 'hat'), 12) |
|
264
|
|
|
0.777777777778 |
|
265
|
|
|
>>> round(sim_jaro_winkler('Niall', 'Neil'), 12) |
|
266
|
|
|
0.805 |
|
267
|
|
|
>>> round(sim_jaro_winkler('aluminum', 'Catalan'), 12) |
|
268
|
|
|
0.60119047619 |
|
269
|
|
|
>>> round(sim_jaro_winkler('ATCG', 'TAGC'), 12) |
|
270
|
|
|
0.833333333333 |
|
271
|
|
|
|
|
272
|
|
|
>>> round(sim_jaro_winkler('cat', 'hat', mode='jaro'), 12) |
|
273
|
|
|
0.777777777778 |
|
274
|
|
|
>>> round(sim_jaro_winkler('Niall', 'Neil', mode='jaro'), 12) |
|
275
|
|
|
0.783333333333 |
|
276
|
|
|
>>> round(sim_jaro_winkler('aluminum', 'Catalan', mode='jaro'), 12) |
|
277
|
|
|
0.60119047619 |
|
278
|
|
|
>>> round(sim_jaro_winkler('ATCG', 'TAGC', mode='jaro'), 12) |
|
279
|
|
|
0.833333333333 |
|
280
|
|
|
|
|
281
|
|
|
""" |
|
282
|
1 |
|
return JaroWinkler().sim( |
|
283
|
|
|
src, tar, qval, mode, long_strings, boost_threshold, scaling_factor |
|
284
|
|
|
) |
|
285
|
|
|
|
|
286
|
|
|
|
|
287
|
1 |
|
def dist_jaro_winkler( |
|
|
|
|
|
|
288
|
|
|
src, |
|
|
|
|
|
|
289
|
|
|
tar, |
|
|
|
|
|
|
290
|
|
|
qval=1, |
|
|
|
|
|
|
291
|
|
|
mode='winkler', |
|
|
|
|
|
|
292
|
|
|
long_strings=False, |
|
|
|
|
|
|
293
|
|
|
boost_threshold=0.7, |
|
|
|
|
|
|
294
|
|
|
scaling_factor=0.1, |
|
|
|
|
|
|
295
|
|
|
): |
|
296
|
|
|
"""Return the Jaro or Jaro-Winkler distance between two strings. |
|
297
|
|
|
|
|
298
|
|
|
This is a wrapper for :py:meth:`JaroWinkler.dist`. |
|
299
|
|
|
|
|
300
|
|
|
Args: |
|
301
|
|
|
src (str): Source string for comparison |
|
302
|
|
|
tar (str): Target string for comparison |
|
303
|
|
|
qval (int): The length of each q-gram (defaults to 1: |
|
304
|
|
|
character-wise matching) |
|
305
|
|
|
mode (str): Indicates which variant of this distance metric to |
|
306
|
|
|
compute: |
|
307
|
|
|
- ``winkler`` -- computes the Jaro-Winkler distance (default) |
|
308
|
|
|
which increases the score for matches near the start of the |
|
309
|
|
|
word |
|
310
|
|
|
- ``jaro`` -- computes the Jaro distance |
|
311
|
|
|
long_strings (bool): Set to True to "Increase the probability of a |
|
312
|
|
|
match when the number of matched characters is large. This option |
|
313
|
|
|
allows for a little more tolerance when the strings are large. It |
|
314
|
|
|
is not an appropriate test when comparing fixedlength fields such |
|
315
|
|
|
as phone and social security numbers." (Used in 'winkler' mode |
|
316
|
|
|
only.) |
|
317
|
|
|
boost_threshold (float): A value between 0 and 1, below which the |
|
318
|
|
|
Winkler boost is not applied (defaults to 0.7). (Used in 'winkler' |
|
319
|
|
|
mode only.) |
|
320
|
|
|
scaling_factor (float): A value between 0 and 0.25, indicating by how |
|
321
|
|
|
much to boost scores for matching prefixes (defaults to 0.1). (Used |
|
322
|
|
|
in 'winkler' mode only.) |
|
323
|
|
|
|
|
324
|
|
|
Returns: |
|
325
|
|
|
float: Jaro or Jaro-Winkler distance |
|
326
|
|
|
|
|
327
|
|
|
Examples: |
|
328
|
|
|
>>> round(dist_jaro_winkler('cat', 'hat'), 12) |
|
329
|
|
|
0.222222222222 |
|
330
|
|
|
>>> round(dist_jaro_winkler('Niall', 'Neil'), 12) |
|
331
|
|
|
0.195 |
|
332
|
|
|
>>> round(dist_jaro_winkler('aluminum', 'Catalan'), 12) |
|
333
|
|
|
0.39880952381 |
|
334
|
|
|
>>> round(dist_jaro_winkler('ATCG', 'TAGC'), 12) |
|
335
|
|
|
0.166666666667 |
|
336
|
|
|
|
|
337
|
|
|
>>> round(dist_jaro_winkler('cat', 'hat', mode='jaro'), 12) |
|
338
|
|
|
0.222222222222 |
|
339
|
|
|
>>> round(dist_jaro_winkler('Niall', 'Neil', mode='jaro'), 12) |
|
340
|
|
|
0.216666666667 |
|
341
|
|
|
>>> round(dist_jaro_winkler('aluminum', 'Catalan', mode='jaro'), 12) |
|
342
|
|
|
0.39880952381 |
|
343
|
|
|
>>> round(dist_jaro_winkler('ATCG', 'TAGC', mode='jaro'), 12) |
|
344
|
|
|
0.166666666667 |
|
345
|
|
|
|
|
346
|
|
|
""" |
|
347
|
1 |
|
return JaroWinkler().dist( |
|
348
|
|
|
src, tar, qval, mode, long_strings, boost_threshold, scaling_factor |
|
349
|
|
|
) |
|
350
|
|
|
|
|
351
|
|
|
|
|
352
|
|
|
if __name__ == '__main__': |
|
353
|
|
|
import doctest |
|
354
|
|
|
|
|
355
|
|
|
doctest.testmod() |
|
356
|
|
|
|
chrislit /
abydos
Loading history...