normalize.py
10.9 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
# -*- coding: utf-8 -*-
import itertools
import settings
from verification.nkjp import NKJP
from wordnet.models import LexicalUnit, Relation
def expressions(expressions):
orth_expressions = []
nkjp = None
if settings.NKJP_VALIDATION:
nkjp = NKJP('nkjp1800x')
for expr in expressions.order_by('text'):
if expr.segments.count() < 2:
continue
no_spaced_orig_expr = expr.orth_text.replace(' ', '')
for generated in expression(nkjp, expr):
no_spaced_gen_expr = generated.replace(' ', '')
if no_spaced_gen_expr != no_spaced_orig_expr:
orth_expressions.append(generated)
if settings.NKJP_VALIDATION:
nkjp.close()
return orth_expressions
# @TODO: czasowniki z sie przy szukaniu synonimow
def expression(nkjp, expression):
normalized_expressions = []
if expression.segments.count() > 1:
possible_forms = generate_forms(expression)
if len(possible_forms) > 1:
for form in possible_forms:
if (not settings.NKJP_VALIDATION
or nkjp.check_expression(expression, form)):
normalized_expression = get_normalized_expr_text(expression, form)
normalized_expressions.append(normalized_expression)
return set(normalized_expressions)
def generate_forms(expression):
segments = []
expr_segments = expression.segments.order_by('position_in_expr')
for seg in expr_segments:
seg_lus = get_lus(seg.base)
if seg.is_head and (seg_lus.count() == 1 or settings.NKJP_VALIDATION):
head_equivalents = get_head_equivalents(seg)
segments.append(head_equivalents)
elif seg_lus.count() == 1 or settings.NKJP_VALIDATION:
equivalents = get_synonymic_equivalents(seg)
segments.append(equivalents)
else:
segments.append([seg.orth])
generated_expressions = itertools.product(*segments)
generated_expressions = list(generated_expressions)
pariciple_expressions = create_pariciple_expressions(generated_expressions, expression)
generated_expressions.extend(pariciple_expressions)
return generated_expressions
def create_pariciple_expressions(generated_expressions, expression):
if u'który' not in expression.base_text:
return []
pariciple_expressions = []
original_segments = expression.segments.order_by('position_in_expr')
for i, orig_seg in enumerate(original_segments):
ifin = -1
if is_aff_construction(original_segments, i):
gender = get_adj_gender(original_segments[i + 1].msd)
case = get_adj_case(original_segments[i + 1].msd)
if 'nom' in case.split('.'):
ctag = 'pact'
case = 'nom.voc'
ifin = i + 2
elif is_neg_construction(original_segments, i):
gender = get_adj_gender(original_segments[i + 1].msd)
case = get_adj_case(original_segments[i + 1].msd)
if 'nom' in case.split('.'):
ctag = 'pact'
case = 'nom.voc'
ifin = i + 3
if ifin >= 0:
pariciple_expressions.extend(get_participle_expressions(generated_expressions, original_segments, ifin,
gender, case, ctag))
return pariciple_expressions
def is_aff_construction(segments, start):
if (start + 2 < segments.count() and
segments[start].orth == u',' and
segments[start + 1].base == u'który' and
segments[start + 2].ctag == u'fin'):
return True
return False
def is_neg_construction(segments, start):
if (start + 3 < segments.count() and
segments[start].orth == u',' and
segments[start + 1].base == u'który' and
segments[start + 2].base == u'nie' and segments[start + 2].ctag != u'ppron3' and
segments[start + 3].ctag == u'fin'):
return True
return False
def get_participle_expressions(orig_expressions, segments, start, gender, case, ctag):
iseg = start
participle_expressions = []
verbs = []
while iseg < segments.count() and (segments[iseg].ctag in ['interp', 'fin', 'conj'] or segments[iseg].base in [u'się', u'nie']):
if segments[iseg].ctag == 'fin':
number = get_fin_number(segments[iseg].msd)
aspect = get_fin_aspect(segments[iseg].msd)
verb_base = segments[iseg].base
if verb_base == u'być':
return []
negation = 'aff'
if iseg - 1 > 0 and segments[iseg - 1].base == u'nie' and segments[iseg - 1].ctag != u'ppron3':
negation = 'neg'
possible_participles = get_matching_forms(verb_base, ctag,
number, case, gender, aspect, negation)
verbs.append({'seg': segments[iseg],
'iseg': iseg,
'negation': negation,
'possible_participles': possible_participles})
iseg += 1
for expr in orig_expressions:
participle_expr = list(expr)
for verb in verbs:
# for part in verb['possible_participles']:
participle_expr[verb['iseg']] = verb['possible_participles'][0]
if verb['iseg'] == start and verb['negation'] == 'neg':
participle_expr[verb['iseg'] - 3] = None
participle_expr[verb['iseg'] - 2] = None
participle_expr[verb['iseg'] - 1] = None
elif verb['iseg'] == start and verb['negation'] == 'aff':
participle_expr[verb['iseg'] - 2] = None
participle_expr[verb['iseg'] - 1] = None
elif verb['iseg'] != start and verb['negation'] == 'neg':
participle_expr[verb['iseg'] - 1] = None
if (verb['iseg'] + 1 < segments.count() and
segments[verb['iseg'] + 1].base == u'się' and
ctag == 'ppas'):
participle_expr[verb['iseg'] + 1] = None
participle_expressions.append(participle_expr)
return participle_expressions
def get_adj_case(msd):
tags = msd.split(':')
return tags[1]
def get_adj_gender(msd):
tags = msd.split(':')
return tags[2]
def get_fin_number(msd):
tags = msd.split(':')
return tags[0]
def get_fin_aspect(msd):
tags = msd.split(':')
return tags[2]
def get_matching_forms(base, ctag, number, case, gender, aspect, negation):
matching_forms = []
possible_forms = settings.MORFEUSZ2.generate(base.encode('utf8'))
for form in possible_forms:
form_tags = form.getTag(settings.MORFEUSZ2).split(':')
if (ctag == form_tags[0] and
tags_match(form_tags[1], number) and
tags_match(form_tags[2], case) and
tags_match(form_tags[3], gender) and
tags_match(form_tags[4], aspect) and
tags_match(form_tags[5], negation)):
if form.orth not in matching_forms:
matching_forms.append(form.orth)
return matching_forms
def tags_match(tag1, tag2):
tag1_alternatives = tag1.split('.')
tag2_alternatives = tag2.split('.')
return bool(set(tag1_alternatives) & set(tag2_alternatives))
def get_synonymic_equivalents(seg):
all_equivalents = []
all_equivalents.extend(get_synonyms(seg.base))
inflected_equivalents = [seg.orth]
for equivalent in all_equivalents:
forms = segment(seg, equivalent)
inflected_equivalents.extend(forms)
return list(set(inflected_equivalents))
def get_head_equivalents(seg):
all_equivalents = []
all_equivalents.extend(get_synonyms(seg.base))
all_equivalents.extend(get_by_relation(seg.base, 1,
Relation.objects.get(parent=None, name='hiperonimia')))
inflected_equivalents = [seg.orth]
for equivalent in all_equivalents:
forms = segment(seg, equivalent)
inflected_equivalents.extend(forms)
return list(set(inflected_equivalents))
def get_synonyms(base):
synonyms = []
synsets = get_synsets(base)
for synset in synsets:
for lu in synset.lus.all():
if (base != lu.base and len(lu.base.split()) == 1 and lu.base not in synonyms):
synonyms.append(lu.base)
return synonyms
def get_synsets(base):
synsets = []
lus = get_lus(base)
for lu in lus:
if lu.synset not in synsets:
synsets.append(lu.synset)
return synsets
def get_lus(base):
return LexicalUnit.objects.filter(base=base)
def get_by_relation(base, max_depth, relation):
related_bases = []
source_synsets = get_synsets(base)
related_synsets = get_related_synsets(source_synsets, relation, max_depth)
for synset in related_synsets:
if synset not in source_synsets:
for lu in synset.lus.all():
if (base != lu.base and len(lu.base.split()) == 1 and
lu.base not in related_bases):
related_bases.append(lu.base)
return related_bases
def get_related_synsets(sources, relation, max_depth):
related_synsets = []
for source in sources:
related_synsets.extend(get_related(source, relation, max_depth, 0))
return related_synsets
def get_related(source, relation, max_depth, depth):
depth += 1
visited = [source]
if depth > max_depth:
return visited
links = source.targets.filter(relation=relation)
for lid, link in enumerate(links):
visited.extend(get_related(link.parent, relation, max_depth, depth))
return visited
def get_normalized_expr_text(expression, form_segments):
expr = ''
expr_segments = expression.segments.order_by('position_in_expr')
for expr_seg, form_seg in itertools.izip(expr_segments, form_segments):
if form_seg is None:
continue
orth = form_seg
if expr_seg.has_nps:
expr += orth
else:
expr += ' %s' % orth
return expr.lstrip()
def segment(orig_seg, synonym):
return generate_inflected_forms(synonym, orig_seg)
def generate_inflected_forms(synonym, orig_seg):
inflected_forms = []
try:
orig_tag = '%s:%s' % (orig_seg.ctag, orig_seg.msd)
orig_tag_id = settings.MORFEUSZ2.getIdResolver().getTagId(orig_tag)
inflected_forms = inflect(synonym, orig_tag_id)
except RuntimeError:
pass
# orig_morf2_interps = settings.MORFEUSZ2.analyse(orig_seg.orth.encode('utf8'))
# for interp in orig_morf2_interps:
# if interp.getTag(settings.MORFEUSZ2) == u'%s:%s' % (orig_seg.ctag, orig_seg.msd):
# inflected.extend(inflect(synonym, interp.tagId))
return set(inflected_forms)
def inflect(base, tag_id):
possible_forms = settings.MORFEUSZ2.generate(base.encode('utf8'), tag_id)
return [form.orth for form in possible_forms]