|
1
|
|
|
# Copyright 2018-2020 by Christopher C. Little. |
|
2
|
|
|
# This file is part of Abydos. |
|
3
|
|
|
# |
|
4
|
|
|
# Abydos is free software: you can redistribute it and/or modify |
|
5
|
|
|
# it under the terms of the GNU General Public License as published by |
|
6
|
|
|
# the Free Software Foundation, either version 3 of the License, or |
|
7
|
|
|
# (at your option) any later version. |
|
8
|
|
|
# |
|
9
|
|
|
# Abydos is distributed in the hope that it will be useful, |
|
10
|
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
11
|
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
12
|
|
|
# GNU General Public License for more details. |
|
13
|
|
|
# |
|
14
|
|
|
# You should have received a copy of the GNU General Public License |
|
15
|
|
|
# along with Abydos. If not, see <http://www.gnu.org/licenses/>. |
|
16
|
|
|
|
|
17
|
|
|
"""abydos.stemmer._uea_lite. |
|
18
|
|
|
|
|
19
|
1 |
|
UEA-Lite stemmer |
|
20
|
|
|
""" |
|
21
|
|
|
|
|
22
|
|
|
from re import match as re_match |
|
23
|
|
|
from typing import Dict, Optional, Tuple |
|
24
|
1 |
|
|
|
25
|
|
|
from ._stemmer import _Stemmer |
|
26
|
|
|
|
|
27
|
|
|
__all__ = ['UEALite'] |
|
28
|
|
|
|
|
29
|
|
|
|
|
30
|
|
|
class UEALite(_Stemmer): |
|
31
|
1 |
|
"""UEA-Lite stemmer. |
|
32
|
|
|
|
|
33
|
1 |
|
The UEA-Lite stemmer is discussed in :cite:`Jenkins:2005`. |
|
34
|
|
|
|
|
35
|
1 |
|
This is chiefly based on the Java implementation of the algorithm, with |
|
36
|
|
|
variants based on the Perl implementation and Jason Adams' Ruby port. |
|
37
|
1 |
|
|
|
38
|
1 |
|
Java version: :cite:`Churchill:2005` |
|
39
|
|
|
Perl version: :cite:`Jenkins:2005` |
|
40
|
1 |
|
Ruby version: :cite:`Adams:2017` |
|
41
|
|
|
|
|
42
|
|
|
.. versionadded:: 0.3.6 |
|
43
|
1 |
|
""" |
|
44
|
|
|
|
|
45
|
|
|
_problem_words = {'is', 'as', 'this', 'has', 'was', 'during'} |
|
46
|
|
|
|
|
47
|
|
|
# rule table format: |
|
48
|
|
|
# top-level dictionary: length-of-suffix: dict-of-rules |
|
49
|
|
|
# dict-of-rules: suffix: (rule_no, suffix_length_to_delete, |
|
50
|
|
|
# suffix_to_append) |
|
51
|
|
|
_standard_rule_table = { |
|
52
|
|
|
7: { |
|
53
|
|
|
'titudes': (30, 1, None), |
|
54
|
|
|
'fulness': (34, 4, None), |
|
55
|
|
|
'ousness': (35, 4, None), |
|
56
|
|
|
'eadings': (40.7, 4, None), |
|
57
|
|
|
'oadings': (40.6, 4, None), |
|
58
|
1 |
|
'ealings': (42.4, 4, None), |
|
59
|
|
|
'ailings': (42.2, 4, None), |
|
60
|
|
|
}, |
|
61
|
|
|
6: { |
|
62
|
|
|
'aceous': (1, 6, None), |
|
63
|
|
|
'aining': (24, 3, None), |
|
64
|
1 |
|
'acting': (25, 3, None), |
|
65
|
|
|
'ttings': (26, 5, None), |
|
66
|
|
|
'viding': (27, 3, 'e'), |
|
67
|
|
|
'ssings': (37, 4, None), |
|
68
|
|
|
'ulting': (38, 3, None), |
|
69
|
|
|
'eading': (40.7, 3, None), |
|
70
|
|
|
'oading': (40.6, 3, None), |
|
71
|
|
|
'edings': (40.5, 4, None), |
|
72
|
|
|
'ddings': (40.4, 5, None), |
|
73
|
|
|
'ldings': (40.3, 4, None), |
|
74
|
|
|
'rdings': (40.2, 4, None), |
|
75
|
|
|
'ndings': (40.1, 4, None), |
|
76
|
|
|
'llings': (41, 5, None), |
|
77
|
|
|
'ealing': (42.4, 3, None), |
|
78
|
|
|
'olings': (42.3, 4, None), |
|
79
|
|
|
'ailing': (42.2, 3, None), |
|
80
|
|
|
'elings': (42.1, 4, None), |
|
81
|
|
|
'mmings': (44.3, 5, None), |
|
82
|
|
|
'ngings': (45.2, 4, None), |
|
83
|
|
|
'ggings': (45.1, 5, None), |
|
84
|
|
|
'stings': (47, 4, None), |
|
85
|
|
|
'etings': (48.4, 4, None), |
|
86
|
|
|
'ntings': (48.2, 4, None), |
|
87
|
|
|
'irings': (54.4, 4, 'e'), |
|
88
|
|
|
'urings': (54.3, 4, 'e'), |
|
89
|
|
|
'ncings': (54.2, 4, 'e'), |
|
90
|
|
|
'things': (58.1, 1, None), |
|
91
|
|
|
}, |
|
92
|
|
|
5: { |
|
93
|
|
|
'iases': (11.4, 2, None), |
|
94
|
|
|
'ained': (13.6, 2, None), |
|
95
|
|
|
'erned': (13.5, 2, None), |
|
96
|
|
|
'ifted': (14, 2, None), |
|
97
|
|
|
'ected': (15, 2, None), |
|
98
|
|
|
'vided': (16, 1, None), |
|
99
|
|
|
'erred': (19, 3, None), |
|
100
|
|
|
'urred': (20.5, 3, None), |
|
101
|
|
|
'lored': (20.4, 2, None), |
|
102
|
|
|
'eared': (20.3, 2, None), |
|
103
|
|
|
'tored': (20.2, 1, None), |
|
104
|
|
|
'noted': (22.4, 1, None), |
|
105
|
|
|
'leted': (22.3, 1, None), |
|
106
|
|
|
'anges': (23, 1, None), |
|
107
|
|
|
'tting': (26, 4, None), |
|
108
|
|
|
'ulted': (32, 2, None), |
|
109
|
|
|
'uming': (33, 3, 'e'), |
|
110
|
|
|
'rabed': (36.1, 1, None), |
|
111
|
|
|
'rebed': (36.1, 1, None), |
|
112
|
|
|
'ribed': (36.1, 1, None), |
|
113
|
|
|
'robed': (36.1, 1, None), |
|
114
|
|
|
'rubed': (36.1, 1, None), |
|
115
|
|
|
'ssing': (37, 3, None), |
|
116
|
|
|
'vings': (39, 4, 'e'), |
|
117
|
|
|
'eding': (40.5, 3, None), |
|
118
|
|
|
'dding': (40.4, 4, None), |
|
119
|
|
|
'lding': (40.3, 3, None), |
|
120
|
|
|
'rding': (40.2, 3, None), |
|
121
|
|
|
'nding': (40.1, 3, None), |
|
122
|
|
|
'dings': (40, 4, 'e'), |
|
123
|
|
|
'lling': (41, 4, None), |
|
124
|
|
|
'oling': (42.3, 3, None), |
|
125
|
|
|
'eling': (42.1, 3, None), |
|
126
|
|
|
'lings': (42, 4, 'e'), |
|
127
|
|
|
'mming': (44.3, 4, None), |
|
128
|
|
|
'rming': (44.2, 3, None), |
|
129
|
|
|
'lming': (44.1, 3, None), |
|
130
|
|
|
'mings': (44, 4, 'e'), |
|
131
|
|
|
'nging': (45.2, 3, None), |
|
132
|
|
|
'gging': (45.1, 4, None), |
|
133
|
|
|
'gings': (45, 4, 'e'), |
|
134
|
|
|
'aning': (46.6, 3, None), |
|
135
|
|
|
'ening': (46.5, 3, None), |
|
136
|
|
|
'gning': (46.4, 3, None), |
|
137
|
|
|
'nning': (46.3, 4, None), |
|
138
|
|
|
'oning': (46.2, 3, None), |
|
139
|
|
|
'rning': (46.1, 3, None), |
|
140
|
|
|
'sting': (47, 3, None), |
|
141
|
|
|
'eting': (48.4, 3, None), |
|
142
|
|
|
'pting': (48.3, 3, None), |
|
143
|
|
|
'nting': (48.2, 3, None), |
|
144
|
|
|
'cting': (48.1, 3, None), |
|
145
|
|
|
'tings': (48, 4, 'e'), |
|
146
|
|
|
'iring': (54.4, 3, 'e'), |
|
147
|
|
|
'uring': (54.3, 3, 'e'), |
|
148
|
|
|
'ncing': (54.2, 3, 'e'), |
|
149
|
|
|
'sings': (54, 4, 'e'), |
|
150
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
|
151
|
|
|
'ating': (57, 3, 'e'), |
|
152
|
|
|
'thing': (58.1, 0, None), |
|
153
|
|
|
}, |
|
154
|
|
|
4: { |
|
155
|
|
|
'eeds': (7, 1, None), |
|
156
|
|
|
'uses': (11.3, 1, None), |
|
157
|
|
|
'sses': (11.2, 2, None), |
|
158
|
|
|
'eses': (11.1, 2, 'is'), |
|
159
|
|
|
'tled': (12.5, 1, None), |
|
160
|
|
|
'pled': (12.4, 1, None), |
|
161
|
|
|
'bled': (12.3, 1, None), |
|
162
|
|
|
'eled': (12.2, 2, None), |
|
163
|
|
|
'lled': (12.1, 2, None), |
|
164
|
|
|
'ened': (13.7, 2, None), |
|
165
|
|
|
'rned': (13.4, 2, None), |
|
166
|
|
|
'nned': (13.3, 3, None), |
|
167
|
|
|
'oned': (13.2, 2, None), |
|
168
|
|
|
'gned': (13.1, 2, None), |
|
169
|
|
|
'ered': (20.1, 2, None), |
|
170
|
|
|
'reds': (20, 2, None), |
|
171
|
|
|
'tted': (21, 3, None), |
|
172
|
|
|
'uted': (22.2, 1, None), |
|
173
|
|
|
'ated': (22.1, 1, None), |
|
174
|
|
|
'ssed': (28, 2, None), |
|
175
|
|
|
'umed': (31, 1, None), |
|
176
|
|
|
'beds': (36, 3, None), |
|
177
|
|
|
'ving': (39, 3, 'e'), |
|
178
|
|
|
'ding': (40, 3, 'e'), |
|
179
|
|
|
'ling': (42, 3, 'e'), |
|
180
|
|
|
'nged': (43.2, 1, None), |
|
181
|
|
|
'gged': (43.1, 3, None), |
|
182
|
|
|
'ming': (44, 3, 'e'), |
|
183
|
|
|
'ging': (45, 3, 'e'), |
|
184
|
|
|
'ning': (46, 3, 'e'), |
|
185
|
|
|
'ting': (48, 3, 'e'), |
|
186
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
|
187
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
|
188
|
|
|
'zing': (54.1, 3, 'e'), |
|
189
|
|
|
'sing': (54, 3, 'e'), |
|
190
|
|
|
'lves': (60.1, 3, 'f'), |
|
191
|
|
|
'aped': (61.3, 1, None), |
|
192
|
|
|
'uded': (61.2, 1, None), |
|
193
|
|
|
'oded': (61.1, 1, None), |
|
194
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
|
195
|
|
|
'ones': (63.6, 1, None), |
|
196
|
|
|
'izes': (63.5, 1, None), |
|
197
|
|
|
'ures': (63.4, 1, None), |
|
198
|
|
|
'ines': (63.3, 1, None), |
|
199
|
|
|
'ides': (63.2, 1, None), |
|
200
|
|
|
}, |
|
201
|
|
|
3: { |
|
202
|
|
|
'ces': (2, 1, None), |
|
203
|
|
|
'sis': (4, 0, None), |
|
204
|
|
|
'tis': (5, 0, None), |
|
205
|
|
|
'eed': (7, 0, None), |
|
206
|
|
|
'ued': (8, 1, None), |
|
207
|
|
|
'ues': (9, 1, None), |
|
208
|
|
|
'ees': (10, 1, None), |
|
209
|
|
|
'ses': (11, 1, None), |
|
210
|
|
|
'led': (12, 2, None), |
|
211
|
|
|
'ned': (13, 1, None), |
|
212
|
|
|
'ved': (17, 1, None), |
|
213
|
|
|
'ced': (18, 1, None), |
|
214
|
|
|
'red': (20, 1, None), |
|
215
|
|
|
'ted': (22, 2, None), |
|
216
|
|
|
'sed': (29, 1, None), |
|
217
|
|
|
'bed': (36, 2, None), |
|
218
|
|
|
'ged': (43, 1, None), |
|
219
|
|
|
'les': (50, 1, None), |
|
220
|
|
|
'tes': (51, 1, None), |
|
221
|
|
|
'zed': (52, 1, None), |
|
222
|
|
|
'ied': (56, 3, 'y'), |
|
223
|
|
|
'ies': (59, 3, 'y'), |
|
224
|
|
|
'ves': (60, 1, None), |
|
225
|
|
|
'pes': (63.8, 1, None), |
|
226
|
|
|
'mes': (63.7, 1, None), |
|
227
|
|
|
'ges': (63.1, 1, None), |
|
228
|
|
|
'ous': (65, 0, None), |
|
229
|
|
|
'ums': (66, 0, None), |
|
230
|
|
|
}, |
|
231
|
|
|
2: { |
|
232
|
|
|
'cs': (3, 0, None), |
|
233
|
|
|
'ss': (6, 0, None), |
|
234
|
|
|
'es': (63, 2, None), |
|
235
|
|
|
'is': (64, 2, 'e'), |
|
236
|
|
|
'us': (67, 0, None), |
|
237
|
|
|
}, |
|
238
|
|
|
} # type: Dict[int, Dict[str, Tuple[float, int, Optional[str]]]] |
|
239
|
|
|
|
|
240
|
|
|
_perl_rule_table = { |
|
241
|
|
|
7: { |
|
242
|
|
|
'titudes': (30, 1, None), |
|
243
|
|
|
'fulness': (34, 4, None), |
|
244
|
|
|
'ousness': (35, 4, None), |
|
245
|
|
|
}, |
|
246
|
|
|
6: { |
|
247
|
|
|
'aceous': (1, 6, None), |
|
248
|
|
|
'aining': (24, 3, None), |
|
249
|
|
|
'acting': (25, 3, None), |
|
250
|
|
|
'viding': (27, 3, 'e'), |
|
251
|
|
|
'ulting': (38, 3, None), |
|
252
|
|
|
'eading': (40.7, 3, None), |
|
253
|
1 |
|
'oading': (40.6, 3, None), |
|
254
|
|
|
'ealing': (42.4, 3, None), |
|
255
|
|
|
'ailing': (42.2, 3, None), |
|
256
|
|
|
}, |
|
257
|
|
|
5: { |
|
258
|
|
|
'iases': (11.4, 2, None), |
|
259
|
|
|
'ained': (13.6, 2, None), |
|
260
|
|
|
'erned': (13.5, 2, None), |
|
261
|
|
|
'ifted': (14, 2, None), |
|
262
|
|
|
'ected': (15, 2, None), |
|
263
|
|
|
'vided': (16, 1, None), |
|
264
|
|
|
'erred': (19, 3, None), |
|
265
|
|
|
'urred': (20.5, 3, None), |
|
266
|
|
|
'lored': (20.4, 2, None), |
|
267
|
|
|
'eared': (20.3, 2, None), |
|
268
|
|
|
'tored': (20.2, 1, None), |
|
269
|
|
|
'noted': (22.4, 1, None), |
|
270
|
|
|
'leted': (22.3, 1, None), |
|
271
|
|
|
'anges': (23, 1, None), |
|
272
|
|
|
'tting': (26, 4, None), |
|
273
|
|
|
'ulted': (32, 2, None), |
|
274
|
|
|
'uming': (33, 3, 'e'), |
|
275
|
|
|
'rabed': (36.1, 1, None), |
|
276
|
|
|
'rebed': (36.1, 1, None), |
|
277
|
|
|
'ribed': (36.1, 1, None), |
|
278
|
|
|
'robed': (36.1, 1, None), |
|
279
|
|
|
'rubed': (36.1, 1, None), |
|
280
|
|
|
'ssing': (37, 3, None), |
|
281
|
|
|
'eding': (40.5, 3, None), |
|
282
|
|
|
'dding': (40.4, 4, None), |
|
283
|
|
|
'lding': (40.3, 3, None), |
|
284
|
|
|
'rding': (40.2, 3, None), |
|
285
|
|
|
'nding': (40.1, 3, None), |
|
286
|
|
|
'lling': (41, 4, None), |
|
287
|
|
|
'oling': (42.3, 3, None), |
|
288
|
|
|
'eling': (42.1, 3, None), |
|
289
|
|
|
'mming': (44.3, 4, None), |
|
290
|
|
|
'rming': (44.2, 3, None), |
|
291
|
|
|
'lming': (44.1, 3, None), |
|
292
|
|
|
'nging': (45.2, 3, None), |
|
293
|
|
|
'gging': (45.1, 4, None), |
|
294
|
|
|
'aning': (46.6, 3, None), |
|
295
|
|
|
'ening': (46.5, 3, None), |
|
296
|
|
|
'gning': (46.4, 3, None), |
|
297
|
|
|
'nning': (46.3, 4, None), |
|
298
|
|
|
'oning': (46.2, 3, None), |
|
299
|
|
|
'rning': (46.1, 3, None), |
|
300
|
|
|
'sting': (47, 3, None), |
|
301
|
|
|
'eting': (48.4, 3, None), |
|
302
|
|
|
'pting': (48.3, 3, None), |
|
303
|
|
|
'nting': (48.2, 3, None), |
|
304
|
|
|
'cting': (48.1, 3, None), |
|
305
|
|
|
'iring': (54.4, 3, 'e'), |
|
306
|
|
|
'uring': (54.3, 3, 'e'), |
|
307
|
|
|
'ncing': (54.2, 3, 'e'), |
|
308
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
|
309
|
|
|
'ating': (57, 3, 'e'), |
|
310
|
|
|
'thing': (58.1, 0, None), |
|
311
|
|
|
}, |
|
312
|
|
|
4: { |
|
313
|
|
|
'uses': (11.3, 1, None), |
|
314
|
|
|
'sses': (11.2, 2, None), |
|
315
|
|
|
'eses': (11.1, 2, 'is'), |
|
316
|
|
|
'tled': (12.5, 1, None), |
|
317
|
|
|
'pled': (12.4, 1, None), |
|
318
|
|
|
'bled': (12.3, 1, None), |
|
319
|
|
|
'eled': (12.2, 2, None), |
|
320
|
|
|
'lled': (12.1, 2, None), |
|
321
|
|
|
'ened': (13.7, 2, None), |
|
322
|
|
|
'rned': (13.4, 2, None), |
|
323
|
|
|
'nned': (13.3, 3, None), |
|
324
|
|
|
'oned': (13.2, 2, None), |
|
325
|
|
|
'gned': (13.1, 2, None), |
|
326
|
|
|
'ered': (20.1, 2, None), |
|
327
|
|
|
'tted': (21, 3, None), |
|
328
|
|
|
'uted': (22.2, 1, None), |
|
329
|
|
|
'ated': (22.1, 1, None), |
|
330
|
|
|
'ssed': (28, 2, None), |
|
331
|
|
|
'umed': (31, 1, None), |
|
332
|
|
|
'ving': (39, 3, 'e'), |
|
333
|
|
|
'ding': (40, 3, 'e'), |
|
334
|
|
|
'ling': (42, 3, 'e'), |
|
335
|
|
|
'nged': (43.2, 1, None), |
|
336
|
|
|
'gged': (43.1, 3, None), |
|
337
|
|
|
'ming': (44, 3, 'e'), |
|
338
|
|
|
'ging': (45, 3, 'e'), |
|
339
|
|
|
'ning': (46, 3, 'e'), |
|
340
|
|
|
'ting': (48, 3, 'e'), |
|
341
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
|
342
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
|
343
|
|
|
'zing': (54.1, 3, 'e'), |
|
344
|
|
|
'sing': (54, 3, 'e'), |
|
345
|
|
|
'lves': (60.1, 3, 'f'), |
|
346
|
|
|
'aped': (61.3, 1, None), |
|
347
|
|
|
'uded': (61.2, 1, None), |
|
348
|
|
|
'oded': (61.1, 1, None), |
|
349
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
|
350
|
|
|
'ones': (63.6, 1, None), |
|
351
|
|
|
'izes': (63.5, 1, None), |
|
352
|
|
|
'ures': (63.4, 1, None), |
|
353
|
|
|
'ines': (63.3, 1, None), |
|
354
|
|
|
'ides': (63.2, 1, None), |
|
355
|
|
|
}, |
|
356
|
|
|
3: { |
|
357
|
|
|
'ces': (2, 1, None), |
|
358
|
|
|
'sis': (4, 0, None), |
|
359
|
|
|
'tis': (5, 0, None), |
|
360
|
|
|
'eed': (7, 0, None), |
|
361
|
|
|
'ued': (8, 1, None), |
|
362
|
|
|
'ues': (9, 1, None), |
|
363
|
|
|
'ees': (10, 1, None), |
|
364
|
|
|
'ses': (11, 1, None), |
|
365
|
|
|
'led': (12, 2, None), |
|
366
|
|
|
'ned': (13, 1, None), |
|
367
|
|
|
'ved': (17, 1, None), |
|
368
|
|
|
'ced': (18, 1, None), |
|
369
|
|
|
'red': (20, 1, None), |
|
370
|
|
|
'ted': (22, 2, None), |
|
371
|
|
|
'sed': (29, 1, None), |
|
372
|
|
|
'bed': (36, 2, None), |
|
373
|
|
|
'ged': (43, 1, None), |
|
374
|
|
|
'les': (50, 1, None), |
|
375
|
|
|
'tes': (51, 1, None), |
|
376
|
|
|
'zed': (52, 1, None), |
|
377
|
|
|
'ied': (56, 3, 'y'), |
|
378
|
|
|
'ies': (59, 3, 'y'), |
|
379
|
|
|
'ves': (60, 1, None), |
|
380
|
|
|
'pes': (63.8, 1, None), |
|
381
|
|
|
'mes': (63.7, 1, None), |
|
382
|
|
|
'ges': (63.1, 1, None), |
|
383
|
|
|
'ous': (65, 0, None), |
|
384
|
|
|
'ums': (66, 0, None), |
|
385
|
|
|
}, |
|
386
|
|
|
2: { |
|
387
|
|
|
'cs': (3, 0, None), |
|
388
|
|
|
'ss': (6, 0, None), |
|
389
|
|
|
'es': (63, 2, None), |
|
390
|
|
|
'is': (64, 2, 'e'), |
|
391
|
|
|
'us': (67, 0, None), |
|
392
|
|
|
}, |
|
393
|
|
|
} # type: Dict[int, Dict[str, Tuple[float, int, Optional[str]]]] |
|
394
|
|
|
|
|
395
|
|
|
_adams_rule_table = { |
|
396
|
|
|
7: { |
|
397
|
|
|
'titudes': (30, 1, None), |
|
398
|
|
|
'fulness': (34, 4, None), |
|
399
|
|
|
'ousness': (35, 4, None), |
|
400
|
|
|
'eadings': (40.7, 4, None), |
|
401
|
|
|
'oadings': (40.6, 4, None), |
|
402
|
|
|
'ealings': (42.4, 4, None), |
|
403
|
|
|
'ailings': (42.2, 4, None), |
|
404
|
|
|
}, |
|
405
|
|
|
6: { |
|
406
|
|
|
'aceous': (1, 6, None), |
|
407
|
|
|
'aining': (24, 3, None), |
|
408
|
1 |
|
'acting': (25, 3, None), |
|
409
|
|
|
'ttings': (26, 5, None), |
|
410
|
|
|
'viding': (27, 3, 'e'), |
|
411
|
|
|
'ssings': (37, 4, None), |
|
412
|
|
|
'ulting': (38, 3, None), |
|
413
|
|
|
'eading': (40.7, 3, None), |
|
414
|
|
|
'oading': (40.6, 3, None), |
|
415
|
|
|
'edings': (40.5, 4, None), |
|
416
|
|
|
'ddings': (40.4, 5, None), |
|
417
|
|
|
'ldings': (40.3, 4, None), |
|
418
|
|
|
'rdings': (40.2, 4, None), |
|
419
|
|
|
'ndings': (40.1, 4, None), |
|
420
|
|
|
'llings': (41, 5, None), |
|
421
|
|
|
'ealing': (42.4, 3, None), |
|
422
|
|
|
'olings': (42.3, 4, None), |
|
423
|
|
|
'ailing': (42.2, 3, None), |
|
424
|
|
|
'elings': (42.1, 4, None), |
|
425
|
|
|
'mmings': (44.3, 5, None), |
|
426
|
|
|
'ngings': (45.2, 4, None), |
|
427
|
|
|
'ggings': (45.1, 5, None), |
|
428
|
|
|
'stings': (47, 4, None), |
|
429
|
|
|
'etings': (48.4, 4, None), |
|
430
|
|
|
'ntings': (48.2, 4, None), |
|
431
|
|
|
'irings': (54.4, 4, 'e'), |
|
432
|
|
|
'urings': (54.3, 4, 'e'), |
|
433
|
|
|
'ncings': (54.2, 4, 'e'), |
|
434
|
|
|
'things': (58.1, 1, None), |
|
435
|
|
|
'chited': (22.8, 1, None), |
|
436
|
|
|
}, |
|
437
|
|
|
5: { |
|
438
|
|
|
'iases': (11.4, 2, None), |
|
439
|
|
|
'ained': (13.6, 2, None), |
|
440
|
|
|
'erned': (13.5, 2, None), |
|
441
|
|
|
'ifted': (14, 2, None), |
|
442
|
|
|
'ected': (15, 2, None), |
|
443
|
|
|
# 'vided': (16, 1, None), |
|
444
|
|
|
'erred': (19, 3, None), |
|
445
|
|
|
'urred': (20.5, 3, None), |
|
446
|
|
|
'lored': (20.4, 2, None), |
|
447
|
|
|
'eared': (20.3, 2, None), |
|
448
|
|
|
'tored': (20.2, 1, None), |
|
449
|
|
|
'noted': (22.4, 1, None), |
|
450
|
|
|
'leted': (22.3, 1, None), |
|
451
|
|
|
'anges': (23, 1, None), |
|
452
|
|
|
'tting': (26, 4, None), |
|
453
|
|
|
'ulted': (32, 2, None), |
|
454
|
|
|
'uming': (33, 3, 'e'), |
|
455
|
|
|
'rabed': (36.1, 1, None), |
|
456
|
|
|
'rebed': (36.1, 1, None), |
|
457
|
|
|
'ribed': (36.1, 1, None), |
|
458
|
|
|
'robed': (36.1, 1, None), |
|
459
|
|
|
'rubed': (36.1, 1, None), |
|
460
|
|
|
'ssing': (37, 3, None), |
|
461
|
|
|
'vings': (39, 4, 'e'), |
|
462
|
|
|
'eding': (40.5, 3, None), |
|
463
|
|
|
'dding': (40.4, 4, None), |
|
464
|
|
|
'lding': (40.3, 3, None), |
|
465
|
|
|
'rding': (40.2, 3, None), |
|
466
|
|
|
'nding': (40.1, 3, None), |
|
467
|
|
|
'dings': (40, 4, 'e'), |
|
468
|
|
|
'lling': (41, 4, None), |
|
469
|
|
|
'oling': (42.3, 3, None), |
|
470
|
|
|
'eling': (42.1, 3, None), |
|
471
|
|
|
'lings': (42, 4, 'e'), |
|
472
|
|
|
'mming': (44.3, 4, None), |
|
473
|
|
|
'rming': (44.2, 3, None), |
|
474
|
|
|
'lming': (44.1, 3, None), |
|
475
|
|
|
'mings': (44, 4, 'e'), |
|
476
|
|
|
'nging': (45.2, 3, None), |
|
477
|
|
|
'gging': (45.1, 4, None), |
|
478
|
|
|
'gings': (45, 4, 'e'), |
|
479
|
|
|
'aning': (46.6, 3, None), |
|
480
|
|
|
'ening': (46.5, 3, None), |
|
481
|
|
|
'gning': (46.4, 3, None), |
|
482
|
|
|
'nning': (46.3, 4, None), |
|
483
|
|
|
'oning': (46.2, 3, None), |
|
484
|
|
|
'rning': (46.1, 3, None), |
|
485
|
|
|
'sting': (47, 3, None), |
|
486
|
|
|
'eting': (48.4, 3, None), |
|
487
|
|
|
'pting': (48.3, 3, None), |
|
488
|
|
|
'nting': (48.2, 3, None), |
|
489
|
|
|
'cting': (48.1, 3, None), |
|
490
|
|
|
'tings': (48, 4, 'e'), |
|
491
|
|
|
'iring': (54.4, 3, 'e'), |
|
492
|
|
|
'uring': (54.3, 3, 'e'), |
|
493
|
|
|
'ncing': (54.2, 3, 'e'), |
|
494
|
|
|
'sings': (54, 4, 'e'), |
|
495
|
|
|
# 'lling': (55, 3, None), # masked by 41 |
|
496
|
|
|
'ating': (57, 3, 'e'), |
|
497
|
|
|
'thing': (58.1, 0, None), |
|
498
|
|
|
'dying': (58.2, 4, 'ie'), |
|
499
|
|
|
'tying': (58.2, 4, 'ie'), |
|
500
|
|
|
'vited': (22.6, 1, None), |
|
501
|
|
|
'mited': (22.5, 1, None), |
|
502
|
|
|
'vided': (22.9, 1, None), |
|
503
|
|
|
'mided': (22.10, 1, None), |
|
504
|
|
|
'lying': (58.2, 4, 'ie'), |
|
505
|
|
|
'arred': (19.1, 3, None), |
|
506
|
|
|
}, |
|
507
|
|
|
4: { |
|
508
|
|
|
'eeds': (7, 1, None), |
|
509
|
|
|
'uses': (11.3, 1, None), |
|
510
|
|
|
'sses': (11.2, 2, None), |
|
511
|
|
|
'eses': (11.1, 2, 'is'), |
|
512
|
|
|
'tled': (12.5, 1, None), |
|
513
|
|
|
'pled': (12.4, 1, None), |
|
514
|
|
|
'bled': (12.3, 1, None), |
|
515
|
|
|
'eled': (12.2, 2, None), |
|
516
|
|
|
'lled': (12.1, 2, None), |
|
517
|
|
|
'ened': (13.7, 2, None), |
|
518
|
|
|
'rned': (13.4, 2, None), |
|
519
|
|
|
'nned': (13.3, 3, None), |
|
520
|
|
|
'oned': (13.2, 2, None), |
|
521
|
|
|
'gned': (13.1, 2, None), |
|
522
|
|
|
'ered': (20.1, 2, None), |
|
523
|
|
|
'reds': (20, 2, None), |
|
524
|
|
|
'tted': (21, 3, None), |
|
525
|
|
|
'uted': (22.2, 1, None), |
|
526
|
|
|
'ated': (22.1, 1, None), |
|
527
|
|
|
'ssed': (28, 2, None), |
|
528
|
|
|
'umed': (31, 1, None), |
|
529
|
|
|
'beds': (36, 3, None), |
|
530
|
|
|
'ving': (39, 3, 'e'), |
|
531
|
|
|
'ding': (40, 3, 'e'), |
|
532
|
|
|
'ling': (42, 3, 'e'), |
|
533
|
|
|
'nged': (43.2, 1, None), |
|
534
|
|
|
'gged': (43.1, 3, None), |
|
535
|
|
|
'ming': (44, 3, 'e'), |
|
536
|
|
|
'ging': (45, 3, 'e'), |
|
537
|
|
|
'ning': (46, 3, 'e'), |
|
538
|
|
|
'ting': (48, 3, 'e'), |
|
539
|
|
|
# 'ssed': (49, 2, None), # masked by 28 |
|
540
|
|
|
# 'lled': (53, 2, None), # masked by 12.1 |
|
541
|
|
|
'zing': (54.1, 3, 'e'), |
|
542
|
|
|
'sing': (54, 3, 'e'), |
|
543
|
|
|
'lves': (60.1, 3, 'f'), |
|
544
|
|
|
'aped': (61.3, 1, None), |
|
545
|
|
|
'uded': (61.2, 1, None), |
|
546
|
|
|
'oded': (61.1, 1, None), |
|
547
|
|
|
# 'ated': (61, 1, None), # masked by 22.1 |
|
548
|
|
|
'ones': (63.6, 1, None), |
|
549
|
|
|
'izes': (63.5, 1, None), |
|
550
|
|
|
'ures': (63.4, 1, None), |
|
551
|
|
|
'ines': (63.3, 1, None), |
|
552
|
|
|
'ides': (63.2, 1, None), |
|
553
|
|
|
'ited': (22.7, 2, None), |
|
554
|
|
|
'oked': (31.1, 1, None), |
|
555
|
|
|
'aked': (31.1, 1, None), |
|
556
|
|
|
'iked': (31.1, 1, None), |
|
557
|
|
|
'uked': (31.1, 1, None), |
|
558
|
|
|
'amed': (31, 1, None), |
|
559
|
|
|
'imed': (31, 1, None), |
|
560
|
|
|
'does': (31.2, 2, None), |
|
561
|
|
|
}, |
|
562
|
|
|
3: { |
|
563
|
|
|
'ces': (2, 1, None), |
|
564
|
|
|
'sis': (4, 0, None), |
|
565
|
|
|
'tis': (5, 0, None), |
|
566
|
|
|
'eed': (7, 0, None), |
|
567
|
|
|
'ued': (8, 1, None), |
|
568
|
|
|
'ues': (9, 1, None), |
|
569
|
|
|
'ees': (10, 1, None), |
|
570
|
|
|
'ses': (11, 1, None), |
|
571
|
|
|
'led': (12, 2, None), |
|
572
|
|
|
'ned': (13, 1, None), |
|
573
|
|
|
'ved': (17, 1, None), |
|
574
|
|
|
'ced': (18, 1, None), |
|
575
|
|
|
'red': (20, 1, None), |
|
576
|
|
|
'ted': (22, 2, None), |
|
577
|
|
|
'sed': (29, 1, None), |
|
578
|
|
|
'bed': (36, 2, None), |
|
579
|
|
|
'ged': (43, 1, None), |
|
580
|
|
|
'les': (50, 1, None), |
|
581
|
|
|
'tes': (51, 1, None), |
|
582
|
|
|
'zed': (52, 1, None), |
|
583
|
|
|
'ied': (56, 3, 'y'), |
|
584
|
|
|
'ies': (59, 3, 'y'), |
|
585
|
|
|
'ves': (60, 1, None), |
|
586
|
|
|
'pes': (63.8, 1, None), |
|
587
|
|
|
'mes': (63.7, 1, None), |
|
588
|
|
|
'ges': (63.1, 1, None), |
|
589
|
|
|
'ous': (65, 0, None), |
|
590
|
|
|
'ums': (66, 0, None), |
|
591
|
|
|
'oed': (31.3, 1, None), |
|
592
|
|
|
'oes': (31.2, 1, None), |
|
593
|
|
|
'kes': (63.1, 1, None), |
|
594
|
|
|
'des': (63.10, 1, None), |
|
595
|
|
|
'res': (63.9, 1, None), |
|
596
|
|
|
}, |
|
597
|
|
|
2: { |
|
598
|
|
|
'cs': (3, 0, None), |
|
599
|
|
|
'ss': (6, 0, None), |
|
600
|
|
|
'es': (63, 2, None), |
|
601
|
|
|
'is': (64, 2, 'e'), |
|
602
|
|
|
'us': (67, 0, None), |
|
603
|
|
|
}, |
|
604
|
|
|
} # type: Dict[int, Dict[str, Tuple[float, int, Optional[str]]]] |
|
605
|
|
|
|
|
606
|
|
|
_rules = { |
|
607
|
|
|
'standard': _standard_rule_table, |
|
608
|
|
|
'Adams': _adams_rule_table, |
|
609
|
|
|
'Perl': _perl_rule_table, |
|
610
|
|
|
} # type: Dict[str, Dict[int, Dict[str, Tuple[float, int, Optional[str]]]]] |
|
611
|
|
|
|
|
612
|
|
|
def __init__( |
|
613
|
|
|
self, |
|
614
|
|
|
max_word_length: int = 20, |
|
615
|
|
|
max_acro_length: int = 8, |
|
616
|
|
|
var: str = 'standard', |
|
617
|
|
|
) -> None: |
|
618
|
|
|
"""Initialize UEALite instance. |
|
619
|
1 |
|
|
|
620
|
|
|
Parameters |
|
621
|
|
|
---------- |
|
622
|
|
|
max_word_length : int |
|
623
|
|
|
The maximum word length allowed |
|
624
|
|
|
max_acro_length : int |
|
625
|
1 |
|
The maximum acronym length allowed |
|
626
|
|
|
var : str |
|
627
|
|
|
Variant rules to use: |
|
628
|
|
|
|
|
629
|
|
|
- ``standard`` to use the original (Java-version) rules |
|
630
|
|
|
- ``Adams`` to use Jason Adams' rules |
|
631
|
|
|
- ``Perl`` to use the original Perl rules |
|
632
|
|
|
|
|
633
|
|
|
|
|
634
|
|
|
.. versionadded:: 0.4.0 |
|
635
|
|
|
|
|
636
|
|
|
""" |
|
637
|
|
|
self._max_word_length = max_word_length |
|
638
|
|
|
self._max_acro_length = max_acro_length |
|
639
|
|
|
self._var = var |
|
640
|
|
|
|
|
641
|
|
|
def stem(self, word: str) -> str: |
|
642
|
|
|
"""Return UEA-Lite stem. |
|
643
|
|
|
|
|
644
|
|
|
Parameters |
|
645
|
|
|
---------- |
|
646
|
|
|
word : str |
|
647
|
|
|
The word to stem |
|
648
|
|
|
|
|
649
|
|
|
Returns |
|
650
|
|
|
------- |
|
651
|
|
|
str or (str, int) |
|
652
|
|
|
Word stem |
|
653
|
1 |
|
|
|
654
|
1 |
|
Examples |
|
655
|
1 |
|
-------- |
|
656
|
1 |
|
>>> stmr = UEALite() |
|
657
|
|
|
>>> stmr.stem('readings') |
|
658
|
1 |
|
'read' |
|
659
|
|
|
>>> stmr.stem('insulted') |
|
660
|
|
|
'insult' |
|
661
|
|
|
>>> stmr.stem('cussed') |
|
662
|
|
|
'cuss' |
|
663
|
|
|
>>> stmr.stem('fancies') |
|
664
|
|
|
'fancy' |
|
665
|
|
|
>>> stmr.stem('eroded') |
|
666
|
|
|
'erode' |
|
667
|
|
|
|
|
668
|
|
|
|
|
669
|
|
|
.. versionadded:: 0.1.0 |
|
670
|
|
|
.. versionchanged:: 0.3.6 |
|
671
|
|
|
Encapsulated in class |
|
672
|
|
|
.. versionchanged:: 0.6.0 |
|
673
|
|
|
Made return a str exclusively |
|
674
|
|
|
|
|
675
|
|
|
""" |
|
676
|
|
|
return self._stem_and_rule(word)[0] |
|
677
|
|
|
|
|
678
|
|
|
def _stem_and_rule(self, word: str) -> Tuple[str, float]: |
|
679
|
|
|
"""Return UEA-Lite stem. |
|
680
|
|
|
|
|
681
|
|
|
Parameters |
|
682
|
|
|
---------- |
|
683
|
|
|
word : str |
|
684
|
|
|
The word to stem |
|
685
|
|
|
|
|
686
|
|
|
Returns |
|
687
|
|
|
------- |
|
688
|
|
|
(str, float) |
|
689
|
|
|
Word stem |
|
690
|
|
|
|
|
691
|
1 |
|
Examples |
|
692
|
1 |
|
-------- |
|
693
|
1 |
|
>>> stmr = UEALite() |
|
694
|
1 |
|
>>> stmr._stem_and_rule('readings') |
|
695
|
1 |
|
('read', 40.7) |
|
696
|
1 |
|
>>> stmr._stem_and_rule('insulted') |
|
697
|
1 |
|
('insult', 32) |
|
698
|
|
|
>>> stmr._stem_and_rule('cussed') |
|
699
|
1 |
|
('cuss', 28) |
|
700
|
1 |
|
>>> stmr._stem_and_rule('fancies') |
|
701
|
1 |
|
('fancy', 59) |
|
702
|
|
|
>>> stmr._stem_and_rule('eroded') |
|
703
|
1 |
|
('erode', 61.1) |
|
704
|
1 |
|
|
|
705
|
1 |
|
|
|
706
|
|
|
.. versionadded:: 0.6.0 |
|
707
|
|
|
|
|
708
|
1 |
|
""" |
|
709
|
1 |
|
|
|
710
|
1 |
|
def _stem_with_duplicate_character_check( |
|
711
|
|
|
word: str, del_len: int |
|
712
|
1 |
|
) -> str: |
|
713
|
1 |
|
if word[-1] == 's': |
|
714
|
1 |
|
del_len += 1 |
|
715
|
1 |
|
stemmed_word = word[:-del_len] |
|
716
|
1 |
|
if re_match(r'.*(\w)\1$', stemmed_word): |
|
717
|
1 |
|
stemmed_word = stemmed_word[:-1] |
|
718
|
1 |
|
return stemmed_word |
|
719
|
1 |
|
|
|
720
|
1 |
|
def _stem(word: str) -> Tuple[str, float]: |
|
721
|
1 |
|
stemmed_word = word |
|
722
|
|
|
rule_no = 0.0 |
|
723
|
1 |
|
|
|
724
|
1 |
|
if not word: |
|
725
|
|
|
return word, 0 |
|
726
|
1 |
|
if word in self._problem_words or ( |
|
727
|
1 |
|
word == 'menses' and self._var == 'Adams' |
|
728
|
1 |
|
): |
|
729
|
|
|
return word, 90 |
|
730
|
|
|
if self._max_word_length and len(word) > self._max_word_length: |
|
731
|
|
|
return word, 95 |
|
732
|
1 |
|
|
|
733
|
|
|
if "'" in word: |
|
734
|
1 |
|
if word[-2:] in {"'s", "'S"}: |
|
735
|
1 |
|
stemmed_word = word[:-2] |
|
736
|
1 |
|
if word[-1:] == "'": |
|
737
|
1 |
|
stemmed_word = word[:-1] |
|
738
|
1 |
|
stemmed_word = stemmed_word.replace("n't", 'not') |
|
739
|
|
|
stemmed_word = stemmed_word.replace("'ve", 'have') |
|
740
|
|
|
stemmed_word = stemmed_word.replace("'re", 'are') |
|
741
|
|
|
stemmed_word = stemmed_word.replace("'m", 'am') |
|
742
|
1 |
|
return stemmed_word, 94 |
|
743
|
1 |
|
|
|
744
|
1 |
|
if word.isdigit(): |
|
745
|
1 |
|
return word, 90.3 |
|
746
|
|
|
else: |
|
747
|
|
|
hyphen = word.find('-') |
|
748
|
|
|
if len(word) > hyphen > 0: |
|
749
|
1 |
|
if ( |
|
750
|
1 |
|
word[:hyphen].isalpha() |
|
751
|
1 |
|
and word[hyphen + 1 :].isalpha() |
|
752
|
1 |
|
): |
|
753
|
1 |
|
return word, 90.2 |
|
754
|
1 |
|
else: |
|
755
|
1 |
|
return word, 90.1 |
|
756
|
|
|
elif '_' in word: |
|
757
|
|
|
return word, 90 |
|
758
|
1 |
|
elif word[-1] == 's' and word[:-1].isupper(): |
|
759
|
|
|
if ( |
|
760
|
1 |
|
self._var == 'Adams' |
|
761
|
1 |
|
and len(word) - 1 > self._max_acro_length |
|
762
|
1 |
|
): |
|
763
|
|
|
return word, 96 |
|
764
|
|
|
return word[:-1], 91.1 |
|
765
|
1 |
|
elif word.isupper(): |
|
766
|
1 |
|
if ( |
|
767
|
|
|
self._var == 'Adams' |
|
768
|
1 |
|
and len(word) > self._max_acro_length |
|
769
|
1 |
|
): |
|
770
|
1 |
|
return word, 96 |
|
771
|
1 |
|
return word, 91 |
|
772
|
|
|
elif re_match(r'^.*[A-Z].*[A-Z].*$', word): |
|
773
|
1 |
|
return word, 92 |
|
774
|
1 |
|
elif word[0].isupper(): |
|
775
|
1 |
|
return word, 93 |
|
776
|
|
|
elif self._var == 'Adams' and re_match( |
|
777
|
|
|
r'^[a-z](|[rl])(ing|ed)$', word |
|
778
|
1 |
|
): |
|
779
|
1 |
|
return word, 97 |
|
780
|
1 |
|
|
|
781
|
|
|
for n in range(7, 1, -1): |
|
782
|
|
|
if word[-n:] in self._rules[self._var][n]: |
|
783
|
1 |
|
rule_no, del_len, add_str = self._rules[self._var][n][ |
|
784
|
1 |
|
word[-n:] |
|
785
|
1 |
|
] |
|
786
|
1 |
|
if del_len: |
|
787
|
|
|
stemmed_word = word[:-del_len] |
|
788
|
1 |
|
else: |
|
789
|
|
|
stemmed_word = word |
|
790
|
1 |
|
if add_str: |
|
791
|
1 |
|
stemmed_word += add_str |
|
792
|
1 |
|
break |
|
793
|
1 |
|
|
|
794
|
|
|
if not rule_no: |
|
795
|
|
|
if re_match(r'.*\w\wings?$', word): # rule 58 |
|
796
|
1 |
|
stemmed_word = _stem_with_duplicate_character_check( |
|
797
|
|
|
word, 3 |
|
798
|
|
|
) |
|
799
|
|
|
rule_no = 58 |
|
800
|
|
|
elif re_match(r'.*\w\weds?$', word): # rule 62 |
|
801
|
|
|
stemmed_word = _stem_with_duplicate_character_check( |
|
802
|
1 |
|
word, 2 |
|
803
|
|
|
) |
|
804
|
|
|
rule_no = 62 |
|
805
|
|
|
elif word[-1] == 's': # rule 68 |
|
806
|
|
|
stemmed_word = word[:-1] |
|
807
|
|
|
rule_no = 68 |
|
808
|
|
|
|
|
809
|
|
|
return stemmed_word, rule_no |
|
810
|
|
|
|
|
811
|
|
|
stem, rule_no = _stem(word) |
|
812
|
|
|
|
|
813
|
|
|
return stem, rule_no |
|
814
|
|
|
|
|
815
|
|
|
|
|
816
|
|
|
if __name__ == '__main__': |
|
817
|
|
|
import doctest |
|
818
|
|
|
|
|
819
|
|
|
doctest.testmod() |
|
820
|
|
|
|
chrislit /
abydos
