1
|
|
|
# -*- coding: utf-8 -*- |
2
|
|
|
|
3
|
|
|
# Copyright 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.stemmer._uea_lite. |
20
|
|
|
|
21
|
|
|
UEA-Lite stemmer |
22
|
|
|
""" |
23
|
|
|
|
24
|
1 |
|
from __future__ import ( |
25
|
|
|
absolute_import, |
26
|
|
|
division, |
27
|
|
|
print_function, |
28
|
|
|
unicode_literals, |
29
|
|
|
) |
30
|
|
|
|
31
|
1 |
|
from re import match as re_match |
32
|
|
|
|
33
|
1 |
|
from six.moves import range |
34
|
|
|
|
35
|
1 |
|
from ._stemmer import _Stemmer |
36
|
|
|
|
37
|
1 |
|
__all__ = ['UEALite', 'uealite'] |
38
|
|
|
|
39
|
|
|
|
40
|
1 |
|
class UEALite(_Stemmer): |
|
|
|
|
41
|
|
|
"""UEA-Lite stemmer. |
42
|
|
|
|
43
|
|
|
The UEA-Lite stemmer is discussed in :cite:`Jenkins:2005`. |
44
|
|
|
|
45
|
|
|
This is chiefly based on the Java implementation of the algorithm, with |
46
|
|
|
variants based on the Perl implementation and Jason Adams' Ruby port. |
47
|
|
|
|
48
|
|
|
Java version: :cite:`Churchill:2005` |
49
|
|
|
Perl version: :cite:`Jenkins:2005` |
50
|
|
|
Ruby version: :cite:`Adams:2017` |
51
|
|
|
""" |
52
|
|
|
|
53
|
1 |
|
_problem_words = {'is', 'as', 'this', 'has', 'was', 'during'} |
54
|
|
|
|
55
|
|
|
# rule table format: |
56
|
|
|
# top-level dictionary: length-of-suffix: dict-of-rules |
57
|
|
|
# dict-of-rules: suffix: (rule_no, suffix_length_to_delete, |
58
|
|
|
# suffix_to_append) |
59
|
1 |
|
_standard_rule_table = { |
60
|
|
|
7: { |
61
|
|
|
'titudes': (30, 1, None), |
62
|
|
|
'fulness': (34, 4, None), |
63
|
|
|
'ousness': (35, 4, None), |
64
|
|
|
'eadings': (40.7, 4, None), |
65
|
|
|
'oadings': (40.6, 4, None), |
66
|
|
|
'ealings': (42.4, 4, None), |
67
|
|
|
'ailings': (42.2, 4, None), |
68
|
|
|
}, |
69
|
|
|
6: { |
70
|
|
|
'aceous': (1, 6, None), |
71
|
|
|
'aining': (24, 3, None), |
72
|
|
|
'acting': (25, 3, None), |
73
|
|
|
'ttings': (26, 5, None), |
74
|
|
|
'viding': (27, 3, 'e'), |
75
|
|
|
'ssings': (37, 4, None), |
76
|
|
|
'ulting': (38, 3, None), |
77
|
|
|
'eading': (40.7, 3, None), |
78
|
|
|
'oading': (40.6, 3, None), |
79
|
|
|
'edings': (40.5, 4, None), |
80
|
|
|
'ddings': (40.4, 5, None), |
81
|
|
|
'ldings': (40.3, 4, None), |
82
|
|
|
'rdings': (40.2, 4, None), |
83
|
|
|
'ndings': (40.1, 4, None), |
84
|
|
|
'llings': (41, 5, None), |
85
|
|
|
'ealing': (42.4, 3, None), |
86
|
|
|
'olings': (42.3, 4, None), |
87
|
|
|
'ailing': (42.2, 3, None), |
88
|
|
|
'elings': (42.1, 4, None), |
89
|
|
|
'mmings': (44.3, 5, None), |
90
|
|
|
'ngings': (45.2, 4, None), |
91
|
|
|
'ggings': (45.1, 5, None), |
92
|
|
|
'stings': (47, 4, None), |
93
|
|
|
'etings': (48.4, 4, None), |
94
|
|
|
'ntings': (48.2, 4, None), |
95
|
|
|
'irings': (54.4, 4, 'e'), |
96
|
|
|
'urings': (54.3, 4, 'e'), |
97
|
|
|
'ncings': (54.2, 4, 'e'), |
98
|
|
|
'things': (58.1, 1, None), |
99
|
|
|
}, |
100
|
|
|
5: { |
101
|
|
|
'iases': (11.4, 2, None), |
102
|
|
|
'ained': (13.6, 2, None), |
103
|
|
|
'erned': (13.5, 2, None), |
104
|
|
|
'ifted': (14, 2, None), |
105
|
|
|
'ected': (15, 2, None), |
106
|
|
|
'vided': (16, 1, None), |
107
|
|
|
'erred': (19, 3, None), |
108
|
|
|
'urred': (20.5, 3, None), |
109
|
|
|
'lored': (20.4, 2, None), |
110
|
|
|
'eared': (20.3, 2, None), |
111
|
|
|
'tored': (20.2, 1, None), |
112
|
|
|
'noted': (22.4, 1, None), |
113
|
|
|
'leted': (22.3, 1, None), |
114
|
|
|
'anges': (23, 1, None), |
115
|
|
|
'tting': (26, 4, None), |
116
|
|
|
'ulted': (32, 2, None), |
117
|
|
|
'uming': (33, 3, 'e'), |
118
|
|
|
'rabed': (36.1, 1, None), |
119
|
|
|
'rebed': (36.1, 1, None), |
120
|
|
|
'ribed': (36.1, 1, None), |
121
|
|
|
'robed': (36.1, 1, None), |
122
|
|
|
'rubed': (36.1, 1, None), |
123
|
|
|
'ssing': (37, 3, None), |
124
|
|
|
'vings': (39, 4, 'e'), |
125
|
|
|
'eding': (40.5, 3, None), |
126
|
|
|
'dding': (40.4, 4, None), |
127
|
|
|
'lding': (40.3, 3, None), |
128
|
|
|
'rding': (40.2, 3, None), |
129
|
|
|
'nding': (40.1, 3, None), |
130
|
|
|
'dings': (40, 4, 'e'), |
131
|
|
|
'lling': (41, 4, None), |
132
|
|
|
'oling': (42.3, 3, None), |
133
|
|
|
'eling': (42.1, 3, None), |
134
|
|
|
'lings': (42, 4, 'e'), |
135
|
|
|
'mming': (44.3, 4, None), |
136
|
|
|
'rming': (44.2, 3, None), |
137
|
|
|
'lming': (44.1, 3, None), |
138
|
|
|
'mings': (44, 4, 'e'), |
139
|
|
|
'nging': (45.2, 3, None), |
140
|
|
|
'gging': (45.1, 4, None), |
141
|
|
|
'gings': (45, 4, 'e'), |
142
|
|
|
'aning': (46.6, 3, None), |
143
|
|
|
'ening': (46.5, 3, None), |
144
|
|
|
'gning': (46.4, 3, None), |
145
|
|
|
'nning': (46.3, 4, None), |
146
|
|
|
'oning': (46.2, 3, None), |
147
|
|
|
'rning': (46.1, 3, None), |
148
|
|
|
'sting': (47, 3, None), |
149
|
|
|
'eting': (48.4, 3, None), |
150
|
|
|
'pting': (48.3, 3, None), |
151
|
|
|
'nting': (48.2, 3, None), |
152
|
|
|
'cting': (48.1, 3, None), |
153
|
|
|
'tings': (48, 4, 'e'), |
154
|
|
|
'iring': (54.4, 3, 'e'), |
155
|
|
|
'uring': (54.3, 3, 'e'), |
156
|
|
|
'ncing': (54.2, 3, 'e'), |
157
|
|
|
'sings': (54, 4, 'e'), |
158
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
159
|
|
|
'ating': (57, 3, 'e'), |
160
|
|
|
'thing': (58.1, 0, None), |
161
|
|
|
}, |
162
|
|
|
4: { |
163
|
|
|
'eeds': (7, 1, None), |
164
|
|
|
'uses': (11.3, 1, None), |
165
|
|
|
'sses': (11.2, 2, None), |
166
|
|
|
'eses': (11.1, 2, 'is'), |
167
|
|
|
'tled': (12.5, 1, None), |
168
|
|
|
'pled': (12.4, 1, None), |
169
|
|
|
'bled': (12.3, 1, None), |
170
|
|
|
'eled': (12.2, 2, None), |
171
|
|
|
'lled': (12.1, 2, None), |
172
|
|
|
'ened': (13.7, 2, None), |
173
|
|
|
'rned': (13.4, 2, None), |
174
|
|
|
'nned': (13.3, 3, None), |
175
|
|
|
'oned': (13.2, 2, None), |
176
|
|
|
'gned': (13.1, 2, None), |
177
|
|
|
'ered': (20.1, 2, None), |
178
|
|
|
'reds': (20, 2, None), |
179
|
|
|
'tted': (21, 3, None), |
180
|
|
|
'uted': (22.2, 1, None), |
181
|
|
|
'ated': (22.1, 1, None), |
182
|
|
|
'ssed': (28, 2, None), |
183
|
|
|
'umed': (31, 1, None), |
184
|
|
|
'beds': (36, 3, None), |
185
|
|
|
'ving': (39, 3, 'e'), |
186
|
|
|
'ding': (40, 3, 'e'), |
187
|
|
|
'ling': (42, 3, 'e'), |
188
|
|
|
'nged': (43.2, 1, None), |
189
|
|
|
'gged': (43.1, 3, None), |
190
|
|
|
'ming': (44, 3, 'e'), |
191
|
|
|
'ging': (45, 3, 'e'), |
192
|
|
|
'ning': (46, 3, 'e'), |
193
|
|
|
'ting': (48, 3, 'e'), |
194
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
195
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
196
|
|
|
'zing': (54.1, 3, 'e'), |
197
|
|
|
'sing': (54, 3, 'e'), |
198
|
|
|
'lves': (60.1, 3, 'f'), |
199
|
|
|
'aped': (61.3, 1, None), |
200
|
|
|
'uded': (61.2, 1, None), |
201
|
|
|
'oded': (61.1, 1, None), |
202
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
203
|
|
|
'ones': (63.6, 1, None), |
204
|
|
|
'izes': (63.5, 1, None), |
205
|
|
|
'ures': (63.4, 1, None), |
206
|
|
|
'ines': (63.3, 1, None), |
207
|
|
|
'ides': (63.2, 1, None), |
208
|
|
|
}, |
209
|
|
|
3: { |
210
|
|
|
'ces': (2, 1, None), |
211
|
|
|
'sis': (4, 0, None), |
212
|
|
|
'tis': (5, 0, None), |
213
|
|
|
'eed': (7, 0, None), |
214
|
|
|
'ued': (8, 1, None), |
215
|
|
|
'ues': (9, 1, None), |
216
|
|
|
'ees': (10, 1, None), |
217
|
|
|
'ses': (11, 1, None), |
218
|
|
|
'led': (12, 2, None), |
219
|
|
|
'ned': (13, 1, None), |
220
|
|
|
'ved': (17, 1, None), |
221
|
|
|
'ced': (18, 1, None), |
222
|
|
|
'red': (20, 1, None), |
223
|
|
|
'ted': (22, 2, None), |
224
|
|
|
'sed': (29, 1, None), |
225
|
|
|
'bed': (36, 2, None), |
226
|
|
|
'ged': (43, 1, None), |
227
|
|
|
'les': (50, 1, None), |
228
|
|
|
'tes': (51, 1, None), |
229
|
|
|
'zed': (52, 1, None), |
230
|
|
|
'ied': (56, 3, 'y'), |
231
|
|
|
'ies': (59, 3, 'y'), |
232
|
|
|
'ves': (60, 1, None), |
233
|
|
|
'pes': (63.8, 1, None), |
234
|
|
|
'mes': (63.7, 1, None), |
235
|
|
|
'ges': (63.1, 1, None), |
236
|
|
|
'ous': (65, 0, None), |
237
|
|
|
'ums': (66, 0, None), |
238
|
|
|
}, |
239
|
|
|
2: { |
240
|
|
|
'cs': (3, 0, None), |
241
|
|
|
'ss': (6, 0, None), |
242
|
|
|
'es': (63, 2, None), |
243
|
|
|
'is': (64, 2, 'e'), |
244
|
|
|
'us': (67, 0, None), |
245
|
|
|
}, |
246
|
|
|
} |
247
|
|
|
|
248
|
1 |
|
_perl_rule_table = { |
249
|
|
|
7: { |
250
|
|
|
'titudes': (30, 1, None), |
251
|
|
|
'fulness': (34, 4, None), |
252
|
|
|
'ousness': (35, 4, None), |
253
|
|
|
}, |
254
|
|
|
6: { |
255
|
|
|
'aceous': (1, 6, None), |
256
|
|
|
'aining': (24, 3, None), |
257
|
|
|
'acting': (25, 3, None), |
258
|
|
|
'viding': (27, 3, 'e'), |
259
|
|
|
'ulting': (38, 3, None), |
260
|
|
|
'eading': (40.7, 3, None), |
261
|
|
|
'oading': (40.6, 3, None), |
262
|
|
|
'ealing': (42.4, 3, None), |
263
|
|
|
'ailing': (42.2, 3, None), |
264
|
|
|
}, |
265
|
|
|
5: { |
266
|
|
|
'iases': (11.4, 2, None), |
267
|
|
|
'ained': (13.6, 2, None), |
268
|
|
|
'erned': (13.5, 2, None), |
269
|
|
|
'ifted': (14, 2, None), |
270
|
|
|
'ected': (15, 2, None), |
271
|
|
|
'vided': (16, 1, None), |
272
|
|
|
'erred': (19, 3, None), |
273
|
|
|
'urred': (20.5, 3, None), |
274
|
|
|
'lored': (20.4, 2, None), |
275
|
|
|
'eared': (20.3, 2, None), |
276
|
|
|
'tored': (20.2, 1, None), |
277
|
|
|
'noted': (22.4, 1, None), |
278
|
|
|
'leted': (22.3, 1, None), |
279
|
|
|
'anges': (23, 1, None), |
280
|
|
|
'tting': (26, 4, None), |
281
|
|
|
'ulted': (32, 2, None), |
282
|
|
|
'uming': (33, 3, 'e'), |
283
|
|
|
'rabed': (36.1, 1, None), |
284
|
|
|
'rebed': (36.1, 1, None), |
285
|
|
|
'ribed': (36.1, 1, None), |
286
|
|
|
'robed': (36.1, 1, None), |
287
|
|
|
'rubed': (36.1, 1, None), |
288
|
|
|
'ssing': (37, 3, None), |
289
|
|
|
'eding': (40.5, 3, None), |
290
|
|
|
'dding': (40.4, 4, None), |
291
|
|
|
'lding': (40.3, 3, None), |
292
|
|
|
'rding': (40.2, 3, None), |
293
|
|
|
'nding': (40.1, 3, None), |
294
|
|
|
'lling': (41, 4, None), |
295
|
|
|
'oling': (42.3, 3, None), |
296
|
|
|
'eling': (42.1, 3, None), |
297
|
|
|
'mming': (44.3, 4, None), |
298
|
|
|
'rming': (44.2, 3, None), |
299
|
|
|
'lming': (44.1, 3, None), |
300
|
|
|
'nging': (45.2, 3, None), |
301
|
|
|
'gging': (45.1, 4, None), |
302
|
|
|
'aning': (46.6, 3, None), |
303
|
|
|
'ening': (46.5, 3, None), |
304
|
|
|
'gning': (46.4, 3, None), |
305
|
|
|
'nning': (46.3, 4, None), |
306
|
|
|
'oning': (46.2, 3, None), |
307
|
|
|
'rning': (46.1, 3, None), |
308
|
|
|
'sting': (47, 3, None), |
309
|
|
|
'eting': (48.4, 3, None), |
310
|
|
|
'pting': (48.3, 3, None), |
311
|
|
|
'nting': (48.2, 3, None), |
312
|
|
|
'cting': (48.1, 3, None), |
313
|
|
|
'iring': (54.4, 3, 'e'), |
314
|
|
|
'uring': (54.3, 3, 'e'), |
315
|
|
|
'ncing': (54.2, 3, 'e'), |
316
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
317
|
|
|
'ating': (57, 3, 'e'), |
318
|
|
|
'thing': (58.1, 0, None), |
319
|
|
|
}, |
320
|
|
|
4: { |
321
|
|
|
'uses': (11.3, 1, None), |
322
|
|
|
'sses': (11.2, 2, None), |
323
|
|
|
'eses': (11.1, 2, 'is'), |
324
|
|
|
'tled': (12.5, 1, None), |
325
|
|
|
'pled': (12.4, 1, None), |
326
|
|
|
'bled': (12.3, 1, None), |
327
|
|
|
'eled': (12.2, 2, None), |
328
|
|
|
'lled': (12.1, 2, None), |
329
|
|
|
'ened': (13.7, 2, None), |
330
|
|
|
'rned': (13.4, 2, None), |
331
|
|
|
'nned': (13.3, 3, None), |
332
|
|
|
'oned': (13.2, 2, None), |
333
|
|
|
'gned': (13.1, 2, None), |
334
|
|
|
'ered': (20.1, 2, None), |
335
|
|
|
'tted': (21, 3, None), |
336
|
|
|
'uted': (22.2, 1, None), |
337
|
|
|
'ated': (22.1, 1, None), |
338
|
|
|
'ssed': (28, 2, None), |
339
|
|
|
'umed': (31, 1, None), |
340
|
|
|
'ving': (39, 3, 'e'), |
341
|
|
|
'ding': (40, 3, 'e'), |
342
|
|
|
'ling': (42, 3, 'e'), |
343
|
|
|
'nged': (43.2, 1, None), |
344
|
|
|
'gged': (43.1, 3, None), |
345
|
|
|
'ming': (44, 3, 'e'), |
346
|
|
|
'ging': (45, 3, 'e'), |
347
|
|
|
'ning': (46, 3, 'e'), |
348
|
|
|
'ting': (48, 3, 'e'), |
349
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
350
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
351
|
|
|
'zing': (54.1, 3, 'e'), |
352
|
|
|
'sing': (54, 3, 'e'), |
353
|
|
|
'lves': (60.1, 3, 'f'), |
354
|
|
|
'aped': (61.3, 1, None), |
355
|
|
|
'uded': (61.2, 1, None), |
356
|
|
|
'oded': (61.1, 1, None), |
357
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
358
|
|
|
'ones': (63.6, 1, None), |
359
|
|
|
'izes': (63.5, 1, None), |
360
|
|
|
'ures': (63.4, 1, None), |
361
|
|
|
'ines': (63.3, 1, None), |
362
|
|
|
'ides': (63.2, 1, None), |
363
|
|
|
}, |
364
|
|
|
3: { |
365
|
|
|
'ces': (2, 1, None), |
366
|
|
|
'sis': (4, 0, None), |
367
|
|
|
'tis': (5, 0, None), |
368
|
|
|
'eed': (7, 0, None), |
369
|
|
|
'ued': (8, 1, None), |
370
|
|
|
'ues': (9, 1, None), |
371
|
|
|
'ees': (10, 1, None), |
372
|
|
|
'ses': (11, 1, None), |
373
|
|
|
'led': (12, 2, None), |
374
|
|
|
'ned': (13, 1, None), |
375
|
|
|
'ved': (17, 1, None), |
376
|
|
|
'ced': (18, 1, None), |
377
|
|
|
'red': (20, 1, None), |
378
|
|
|
'ted': (22, 2, None), |
379
|
|
|
'sed': (29, 1, None), |
380
|
|
|
'bed': (36, 2, None), |
381
|
|
|
'ged': (43, 1, None), |
382
|
|
|
'les': (50, 1, None), |
383
|
|
|
'tes': (51, 1, None), |
384
|
|
|
'zed': (52, 1, None), |
385
|
|
|
'ied': (56, 3, 'y'), |
386
|
|
|
'ies': (59, 3, 'y'), |
387
|
|
|
'ves': (60, 1, None), |
388
|
|
|
'pes': (63.8, 1, None), |
389
|
|
|
'mes': (63.7, 1, None), |
390
|
|
|
'ges': (63.1, 1, None), |
391
|
|
|
'ous': (65, 0, None), |
392
|
|
|
'ums': (66, 0, None), |
393
|
|
|
}, |
394
|
|
|
2: { |
395
|
|
|
'cs': (3, 0, None), |
396
|
|
|
'ss': (6, 0, None), |
397
|
|
|
'es': (63, 2, None), |
398
|
|
|
'is': (64, 2, 'e'), |
399
|
|
|
'us': (67, 0, None), |
400
|
|
|
}, |
401
|
|
|
} |
402
|
|
|
|
403
|
1 |
|
_adams_rule_table = { |
404
|
|
|
7: { |
405
|
|
|
'titudes': (30, 1, None), |
406
|
|
|
'fulness': (34, 4, None), |
407
|
|
|
'ousness': (35, 4, None), |
408
|
|
|
'eadings': (40.7, 4, None), |
409
|
|
|
'oadings': (40.6, 4, None), |
410
|
|
|
'ealings': (42.4, 4, None), |
411
|
|
|
'ailings': (42.2, 4, None), |
412
|
|
|
}, |
413
|
|
|
6: { |
414
|
|
|
'aceous': (1, 6, None), |
415
|
|
|
'aining': (24, 3, None), |
416
|
|
|
'acting': (25, 3, None), |
417
|
|
|
'ttings': (26, 5, None), |
418
|
|
|
'viding': (27, 3, 'e'), |
419
|
|
|
'ssings': (37, 4, None), |
420
|
|
|
'ulting': (38, 3, None), |
421
|
|
|
'eading': (40.7, 3, None), |
422
|
|
|
'oading': (40.6, 3, None), |
423
|
|
|
'edings': (40.5, 4, None), |
424
|
|
|
'ddings': (40.4, 5, None), |
425
|
|
|
'ldings': (40.3, 4, None), |
426
|
|
|
'rdings': (40.2, 4, None), |
427
|
|
|
'ndings': (40.1, 4, None), |
428
|
|
|
'llings': (41, 5, None), |
429
|
|
|
'ealing': (42.4, 3, None), |
430
|
|
|
'olings': (42.3, 4, None), |
431
|
|
|
'ailing': (42.2, 3, None), |
432
|
|
|
'elings': (42.1, 4, None), |
433
|
|
|
'mmings': (44.3, 5, None), |
434
|
|
|
'ngings': (45.2, 4, None), |
435
|
|
|
'ggings': (45.1, 5, None), |
436
|
|
|
'stings': (47, 4, None), |
437
|
|
|
'etings': (48.4, 4, None), |
438
|
|
|
'ntings': (48.2, 4, None), |
439
|
|
|
'irings': (54.4, 4, 'e'), |
440
|
|
|
'urings': (54.3, 4, 'e'), |
441
|
|
|
'ncings': (54.2, 4, 'e'), |
442
|
|
|
'things': (58.1, 1, None), |
443
|
|
|
'chited': (22.8, 1, None), |
444
|
|
|
}, |
445
|
|
|
5: { |
446
|
|
|
'iases': (11.4, 2, None), |
447
|
|
|
'ained': (13.6, 2, None), |
448
|
|
|
'erned': (13.5, 2, None), |
449
|
|
|
'ifted': (14, 2, None), |
450
|
|
|
'ected': (15, 2, None), |
451
|
|
|
# 'vided': (16, 1, None), |
452
|
|
|
'erred': (19, 3, None), |
453
|
|
|
'urred': (20.5, 3, None), |
454
|
|
|
'lored': (20.4, 2, None), |
455
|
|
|
'eared': (20.3, 2, None), |
456
|
|
|
'tored': (20.2, 1, None), |
457
|
|
|
'noted': (22.4, 1, None), |
458
|
|
|
'leted': (22.3, 1, None), |
459
|
|
|
'anges': (23, 1, None), |
460
|
|
|
'tting': (26, 4, None), |
461
|
|
|
'ulted': (32, 2, None), |
462
|
|
|
'uming': (33, 3, 'e'), |
463
|
|
|
'rabed': (36.1, 1, None), |
464
|
|
|
'rebed': (36.1, 1, None), |
465
|
|
|
'ribed': (36.1, 1, None), |
466
|
|
|
'robed': (36.1, 1, None), |
467
|
|
|
'rubed': (36.1, 1, None), |
468
|
|
|
'ssing': (37, 3, None), |
469
|
|
|
'vings': (39, 4, 'e'), |
470
|
|
|
'eding': (40.5, 3, None), |
471
|
|
|
'dding': (40.4, 4, None), |
472
|
|
|
'lding': (40.3, 3, None), |
473
|
|
|
'rding': (40.2, 3, None), |
474
|
|
|
'nding': (40.1, 3, None), |
475
|
|
|
'dings': (40, 4, 'e'), |
476
|
|
|
'lling': (41, 4, None), |
477
|
|
|
'oling': (42.3, 3, None), |
478
|
|
|
'eling': (42.1, 3, None), |
479
|
|
|
'lings': (42, 4, 'e'), |
480
|
|
|
'mming': (44.3, 4, None), |
481
|
|
|
'rming': (44.2, 3, None), |
482
|
|
|
'lming': (44.1, 3, None), |
483
|
|
|
'mings': (44, 4, 'e'), |
484
|
|
|
'nging': (45.2, 3, None), |
485
|
|
|
'gging': (45.1, 4, None), |
486
|
|
|
'gings': (45, 4, 'e'), |
487
|
|
|
'aning': (46.6, 3, None), |
488
|
|
|
'ening': (46.5, 3, None), |
489
|
|
|
'gning': (46.4, 3, None), |
490
|
|
|
'nning': (46.3, 4, None), |
491
|
|
|
'oning': (46.2, 3, None), |
492
|
|
|
'rning': (46.1, 3, None), |
493
|
|
|
'sting': (47, 3, None), |
494
|
|
|
'eting': (48.4, 3, None), |
495
|
|
|
'pting': (48.3, 3, None), |
496
|
|
|
'nting': (48.2, 3, None), |
497
|
|
|
'cting': (48.1, 3, None), |
498
|
|
|
'tings': (48, 4, 'e'), |
499
|
|
|
'iring': (54.4, 3, 'e'), |
500
|
|
|
'uring': (54.3, 3, 'e'), |
501
|
|
|
'ncing': (54.2, 3, 'e'), |
502
|
|
|
'sings': (54, 4, 'e'), |
503
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
504
|
|
|
'ating': (57, 3, 'e'), |
505
|
|
|
'thing': (58.1, 0, None), |
506
|
|
|
'dying': (58.2, 4, 'ie'), |
507
|
|
|
'tying': (58.2, 4, 'ie'), |
508
|
|
|
'vited': (22.6, 1, None), |
509
|
|
|
'mited': (22.5, 1, None), |
510
|
|
|
'vided': (22.9, 1, None), |
511
|
|
|
'mided': (22.10, 1, None), |
512
|
|
|
'lying': (58.2, 4, 'ie'), |
513
|
|
|
'arred': (19.1, 3, None), |
514
|
|
|
}, |
515
|
|
|
4: { |
516
|
|
|
'eeds': (7, 1, None), |
517
|
|
|
'uses': (11.3, 1, None), |
518
|
|
|
'sses': (11.2, 2, None), |
519
|
|
|
'eses': (11.1, 2, 'is'), |
520
|
|
|
'tled': (12.5, 1, None), |
521
|
|
|
'pled': (12.4, 1, None), |
522
|
|
|
'bled': (12.3, 1, None), |
523
|
|
|
'eled': (12.2, 2, None), |
524
|
|
|
'lled': (12.1, 2, None), |
525
|
|
|
'ened': (13.7, 2, None), |
526
|
|
|
'rned': (13.4, 2, None), |
527
|
|
|
'nned': (13.3, 3, None), |
528
|
|
|
'oned': (13.2, 2, None), |
529
|
|
|
'gned': (13.1, 2, None), |
530
|
|
|
'ered': (20.1, 2, None), |
531
|
|
|
'reds': (20, 2, None), |
532
|
|
|
'tted': (21, 3, None), |
533
|
|
|
'uted': (22.2, 1, None), |
534
|
|
|
'ated': (22.1, 1, None), |
535
|
|
|
'ssed': (28, 2, None), |
536
|
|
|
'umed': (31, 1, None), |
537
|
|
|
'beds': (36, 3, None), |
538
|
|
|
'ving': (39, 3, 'e'), |
539
|
|
|
'ding': (40, 3, 'e'), |
540
|
|
|
'ling': (42, 3, 'e'), |
541
|
|
|
'nged': (43.2, 1, None), |
542
|
|
|
'gged': (43.1, 3, None), |
543
|
|
|
'ming': (44, 3, 'e'), |
544
|
|
|
'ging': (45, 3, 'e'), |
545
|
|
|
'ning': (46, 3, 'e'), |
546
|
|
|
'ting': (48, 3, 'e'), |
547
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
548
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
549
|
|
|
'zing': (54.1, 3, 'e'), |
550
|
|
|
'sing': (54, 3, 'e'), |
551
|
|
|
'lves': (60.1, 3, 'f'), |
552
|
|
|
'aped': (61.3, 1, None), |
553
|
|
|
'uded': (61.2, 1, None), |
554
|
|
|
'oded': (61.1, 1, None), |
555
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
556
|
|
|
'ones': (63.6, 1, None), |
557
|
|
|
'izes': (63.5, 1, None), |
558
|
|
|
'ures': (63.4, 1, None), |
559
|
|
|
'ines': (63.3, 1, None), |
560
|
|
|
'ides': (63.2, 1, None), |
561
|
|
|
'ited': (22.7, 2, None), |
562
|
|
|
'oked': (31.1, 1, None), |
563
|
|
|
'aked': (31.1, 1, None), |
564
|
|
|
'iked': (31.1, 1, None), |
565
|
|
|
'uked': (31.1, 1, None), |
566
|
|
|
'amed': (31, 1, None), |
567
|
|
|
'imed': (31, 1, None), |
568
|
|
|
'does': (31.2, 2, None), |
569
|
|
|
}, |
570
|
|
|
3: { |
571
|
|
|
'ces': (2, 1, None), |
572
|
|
|
'sis': (4, 0, None), |
573
|
|
|
'tis': (5, 0, None), |
574
|
|
|
'eed': (7, 0, None), |
575
|
|
|
'ued': (8, 1, None), |
576
|
|
|
'ues': (9, 1, None), |
577
|
|
|
'ees': (10, 1, None), |
578
|
|
|
'ses': (11, 1, None), |
579
|
|
|
'led': (12, 2, None), |
580
|
|
|
'ned': (13, 1, None), |
581
|
|
|
'ved': (17, 1, None), |
582
|
|
|
'ced': (18, 1, None), |
583
|
|
|
'red': (20, 1, None), |
584
|
|
|
'ted': (22, 2, None), |
585
|
|
|
'sed': (29, 1, None), |
586
|
|
|
'bed': (36, 2, None), |
587
|
|
|
'ged': (43, 1, None), |
588
|
|
|
'les': (50, 1, None), |
589
|
|
|
'tes': (51, 1, None), |
590
|
|
|
'zed': (52, 1, None), |
591
|
|
|
'ied': (56, 3, 'y'), |
592
|
|
|
'ies': (59, 3, 'y'), |
593
|
|
|
'ves': (60, 1, None), |
594
|
|
|
'pes': (63.8, 1, None), |
595
|
|
|
'mes': (63.7, 1, None), |
596
|
|
|
'ges': (63.1, 1, None), |
597
|
|
|
'ous': (65, 0, None), |
598
|
|
|
'ums': (66, 0, None), |
599
|
|
|
'oed': (31.3, 1, None), |
600
|
|
|
'oes': (31.2, 1, None), |
601
|
|
|
'kes': (63.1, 1, None), |
602
|
|
|
'des': (63.10, 1, None), |
603
|
|
|
'res': (63.9, 1, None), |
604
|
|
|
}, |
605
|
|
|
2: { |
606
|
|
|
'cs': (3, 0, None), |
607
|
|
|
'ss': (6, 0, None), |
608
|
|
|
'es': (63, 2, None), |
609
|
|
|
'is': (64, 2, 'e'), |
610
|
|
|
'us': (67, 0, None), |
611
|
|
|
}, |
612
|
|
|
} |
613
|
|
|
|
614
|
1 |
|
_rules = { |
615
|
|
|
'standard': _standard_rule_table, |
616
|
|
|
'Adams': _adams_rule_table, |
617
|
|
|
'Perl': _perl_rule_table, |
618
|
|
|
} |
619
|
|
|
|
620
|
1 |
|
def stem( |
|
|
|
|
621
|
|
|
self, |
|
|
|
|
622
|
|
|
word, |
|
|
|
|
623
|
|
|
max_word_length=20, |
|
|
|
|
624
|
|
|
max_acro_length=8, |
|
|
|
|
625
|
|
|
return_rule_no=False, |
|
|
|
|
626
|
|
|
var='standard', |
|
|
|
|
627
|
|
|
): |
628
|
|
|
"""Return UEA-Lite stem. |
629
|
|
|
|
630
|
|
|
Parameters |
631
|
|
|
---------- |
632
|
|
|
word : str |
633
|
|
|
The word to stem |
634
|
|
|
max_word_length : int |
635
|
|
|
The maximum word length allowed |
636
|
|
|
max_acro_length : int |
637
|
|
|
The maximum acronym length allowed |
638
|
|
|
return_rule_no : bool |
639
|
|
|
If True, returns the stem along with rule number |
640
|
|
|
var : str |
641
|
|
|
Variant rules to use: |
642
|
|
|
|
643
|
|
|
- ``Adams`` to use Jason Adams' rules |
644
|
|
|
- ``Perl`` to use the original Perl rules |
645
|
|
|
|
646
|
|
|
Returns |
647
|
|
|
------- |
648
|
|
|
str or (str, int) |
649
|
|
|
Word stem |
650
|
|
|
|
651
|
|
|
Examples |
652
|
|
|
-------- |
653
|
|
|
>>> uealite('readings') |
654
|
|
|
'read' |
655
|
|
|
>>> uealite('insulted') |
656
|
|
|
'insult' |
657
|
|
|
>>> uealite('cussed') |
658
|
|
|
'cuss' |
659
|
|
|
>>> uealite('fancies') |
660
|
|
|
'fancy' |
661
|
|
|
>>> uealite('eroded') |
662
|
|
|
'erode' |
663
|
|
|
|
664
|
|
|
""" |
665
|
|
|
|
666
|
1 |
|
def _stem_with_duplicate_character_check(word, del_len): |
667
|
1 |
|
if word[-1] == 's': |
668
|
1 |
|
del_len += 1 |
669
|
1 |
|
stemmed_word = word[:-del_len] |
670
|
1 |
|
if re_match(r'.*(\w)\1$', stemmed_word): |
671
|
1 |
|
stemmed_word = stemmed_word[:-1] |
672
|
1 |
|
return stemmed_word |
673
|
|
|
|
674
|
1 |
|
def _stem(word): |
|
|
|
|
675
|
1 |
|
stemmed_word = word |
676
|
1 |
|
rule_no = 0 |
677
|
|
|
|
678
|
1 |
|
if not word: |
679
|
1 |
|
return word, 0 |
680
|
1 |
|
if word in self._problem_words or ( |
681
|
|
|
word == 'menses' and var == 'Adams' |
|
|
|
|
682
|
|
|
): |
683
|
1 |
|
return word, 90 |
684
|
1 |
|
if max_word_length and len(word) > max_word_length: |
685
|
1 |
|
return word, 95 |
686
|
|
|
|
687
|
1 |
|
if "'" in word: |
688
|
1 |
|
if word[-2:] in {"'s", "'S"}: |
689
|
1 |
|
stemmed_word = word[:-2] |
690
|
1 |
|
if word[-1:] == "'": |
691
|
1 |
|
stemmed_word = word[:-1] |
692
|
1 |
|
stemmed_word = stemmed_word.replace("n't", 'not') |
693
|
1 |
|
stemmed_word = stemmed_word.replace("'ve", 'have') |
694
|
1 |
|
stemmed_word = stemmed_word.replace("'re", 'are') |
695
|
1 |
|
stemmed_word = stemmed_word.replace("'m", 'am') |
696
|
1 |
|
return stemmed_word, 94 |
697
|
|
|
|
698
|
1 |
|
if word.isdigit(): |
699
|
1 |
|
return word, 90.3 |
700
|
|
|
else: |
701
|
1 |
|
hyphen = word.find('-') |
702
|
1 |
|
if len(word) > hyphen > 0: |
703
|
1 |
|
if ( |
|
|
|
|
704
|
|
|
word[:hyphen].isalpha() |
|
|
|
|
705
|
|
|
and word[hyphen + 1 :].isalpha() |
|
|
|
|
706
|
|
|
): |
707
|
1 |
|
return word, 90.2 |
708
|
|
|
else: |
709
|
1 |
|
return word, 90.1 |
710
|
1 |
|
elif '_' in word: |
711
|
1 |
|
return word, 90 |
712
|
1 |
|
elif word[-1] == 's' and word[:-1].isupper(): |
713
|
1 |
|
if var == 'Adams' and len(word) - 1 > max_acro_length: |
714
|
1 |
|
return word, 96 |
715
|
1 |
|
return word[:-1], 91.1 |
716
|
1 |
|
elif word.isupper(): |
717
|
1 |
|
if var == 'Adams' and len(word) > max_acro_length: |
718
|
1 |
|
return word, 96 |
719
|
1 |
|
return word, 91 |
720
|
1 |
|
elif re_match(r'^.*[A-Z].*[A-Z].*$', word): |
721
|
1 |
|
return word, 92 |
722
|
1 |
|
elif word[0].isupper(): |
723
|
1 |
|
return word, 93 |
724
|
1 |
|
elif var == 'Adams' and re_match( |
725
|
|
|
r'^[a-z](|[rl])(ing|ed)$', word |
|
|
|
|
726
|
|
|
): |
727
|
1 |
|
return word, 97 |
728
|
|
|
|
729
|
1 |
|
for n in range(7, 1, -1): |
|
|
|
|
730
|
1 |
|
if word[-n:] in self._rules[var][n]: |
731
|
1 |
|
rule_no, del_len, add_str = self._rules[var][n][word[-n:]] |
732
|
1 |
|
if del_len: |
733
|
1 |
|
stemmed_word = word[:-del_len] |
734
|
|
|
else: |
735
|
1 |
|
stemmed_word = word |
736
|
1 |
|
if add_str: |
737
|
1 |
|
stemmed_word += add_str |
738
|
1 |
|
break |
739
|
|
|
|
740
|
1 |
|
if not rule_no: |
741
|
1 |
|
if re_match(r'.*\w\wings?$', word): # rule 58 |
742
|
1 |
|
stemmed_word = _stem_with_duplicate_character_check( |
743
|
|
|
word, 3 |
744
|
|
|
) |
745
|
1 |
|
rule_no = 58 |
746
|
1 |
|
elif re_match(r'.*\w\weds?$', word): # rule 62 |
747
|
1 |
|
stemmed_word = _stem_with_duplicate_character_check( |
748
|
|
|
word, 2 |
749
|
|
|
) |
750
|
1 |
|
rule_no = 62 |
751
|
1 |
|
elif word[-1] == 's': # rule 68 |
752
|
1 |
|
stemmed_word = word[:-1] |
753
|
1 |
|
rule_no = 68 |
754
|
|
|
|
755
|
1 |
|
return stemmed_word, rule_no |
756
|
|
|
|
757
|
1 |
|
stem, rule_no = _stem(word) |
758
|
1 |
|
if return_rule_no: |
759
|
1 |
|
return stem, rule_no |
760
|
1 |
|
return stem |
761
|
|
|
|
762
|
|
|
|
763
|
1 |
|
def uealite( |
764
|
|
|
word, |
|
|
|
|
765
|
|
|
max_word_length=20, |
|
|
|
|
766
|
|
|
max_acro_length=8, |
|
|
|
|
767
|
|
|
return_rule_no=False, |
|
|
|
|
768
|
|
|
var='standard', |
|
|
|
|
769
|
|
|
): |
770
|
|
|
"""Return UEA-Lite stem. |
771
|
|
|
|
772
|
|
|
This is a wrapper for :py:meth:`UEALite.stem`. |
773
|
|
|
|
774
|
|
|
Parameters |
775
|
|
|
---------- |
776
|
|
|
word : str |
777
|
|
|
The word to stem |
778
|
|
|
max_word_length : int |
779
|
|
|
The maximum word length allowed |
780
|
|
|
max_acro_length : int |
781
|
|
|
The maximum acronym length allowed |
782
|
|
|
return_rule_no : bool |
783
|
|
|
If True, returns the stem along with rule number |
784
|
|
|
var : str |
785
|
|
|
Variant rules to use: |
786
|
|
|
|
787
|
|
|
- ``Adams`` to use Jason Adams' rules |
788
|
|
|
- ``Perl`` to use the original Perl rules |
789
|
|
|
|
790
|
|
|
Returns |
791
|
|
|
------- |
792
|
|
|
str or (str, int) |
793
|
|
|
Word stem |
794
|
|
|
|
795
|
|
|
Examples |
796
|
|
|
-------- |
797
|
|
|
>>> uealite('readings') |
798
|
|
|
'read' |
799
|
|
|
>>> uealite('insulted') |
800
|
|
|
'insult' |
801
|
|
|
>>> uealite('cussed') |
802
|
|
|
'cuss' |
803
|
|
|
>>> uealite('fancies') |
804
|
|
|
'fancy' |
805
|
|
|
>>> uealite('eroded') |
806
|
|
|
'erode' |
807
|
|
|
|
808
|
|
|
""" |
809
|
1 |
|
return UEALite().stem( |
810
|
|
|
word, max_word_length, max_acro_length, return_rule_no, var |
811
|
|
|
) |
812
|
|
|
|
813
|
|
|
|
814
|
|
|
if __name__ == '__main__': |
815
|
|
|
import doctest |
816
|
|
|
|
817
|
|
|
doctest.testmod() |
818
|
|
|
|