File size: 11,166 Bytes
153c03b |
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 |
# -*- coding: utf-8 -*-
"""
Лемматизатор для R&D прототипирования NLP задач в Питоне
25.03.2020 добавлена ефикация в get_lemma2
05.04.2020 добавлено декодирование для частей речи CONJ, PART и PUNCT
"""
from __future__ import division
from __future__ import print_function
import os
import pickle
import pathlib
import gzip
def decode_pos(pos):
if pos in [u'ДЕЕПРИЧАСТИЕ', u'ГЛАГОЛ', u'ИНФИНИТИВ']:
return u'ГЛАГОЛ'
else:
return pos
class Lemmatizer(object):
def __init__(self):
pass
def load(self, dict_path=None):
""" Загружаем модель лемматизации, созданную отдельным скриптом builder.py """
dict_filename = 'rulemma.dat'
if dict_path is None:
module_folder = str(pathlib.Path(__file__).resolve().parent)
p = os.path.join(module_folder, '../tmp', dict_filename)
if not os.path.exists(p):
p = os.path.join(module_folder, dict_filename)
else:
p = dict_path
with gzip.open(p, 'r') as f:
self.forms, self.forms2, self.special_lemmas, self.key2transducer = pickle.load(f)
def get_lemma(self, word):
if word in self.forms:
return self.forms[word]
elif word in self.forms2:
return self.forms2[word][0]
elif word in self.special_lemmas:
return self.special_lemmas[word]
else:
return word
def decode_pos_tags(self, pos_tags):
stags1 = []
part_of_speech = u'unk'
short_tag_index = -1
for tag in pos_tags.split('|'):
if tag == 'NOUN':
part_of_speech = u'СУЩЕСТВИТЕЛЬНОЕ'
elif tag == 'VERB':
part_of_speech = u'ГЛАГОЛ'
elif tag == 'ADJ':
part_of_speech = u'ПРИЛАГАТЕЛЬНОЕ'
stags1.append((u'КРАТКИЙ', u'0'))
short_tag_index = 0
elif tag == 'ADV':
part_of_speech = u'НАРЕЧИЕ'
elif tag == 'PRON':
part_of_speech = u'МЕСТОИМЕНИЕ'
elif tag == 'ADP':
part_of_speech = u'ПРЕДЛОГ'
elif tag == 'CONJ':
part_of_speech = u'СОЮЗ'
elif tag == 'PART':
part_of_speech = u'ЧАСТИЦА'
elif tag == 'PUNCT':
part_of_speech = u'ПУНКТУАТОР'
elif '=' in tag:
if part_of_speech == u'СУЩЕСТВИТЕЛЬНОЕ':
if tag == u'Case=Nom':
stags1.append((u'ПАДЕЖ', u'ИМ'))
elif tag == u'Case=Acc':
stags1.append((u'ПАДЕЖ', u'ВИН'))
elif tag == u'Case=Dat':
stags1.append((u'ПАДЕЖ', u'ДАТ'))
elif tag == u'Case=Ins':
stags1.append((u'ПАДЕЖ', u'ТВОР'))
elif tag == u'Case=Prep':
stags1.append((u'ПАДЕЖ', u'ПРЕДЛ'))
elif tag == u'Case=Loc':
stags1.append((u'ПАДЕЖ', u'ПРЕДЛ')) # 03-02-2020 u'МЕСТ'
elif tag == u'Case=Gen':
stags1.append((u'ПАДЕЖ', u'РОД'))
elif tag == u'Case=Voc':
stags1.append((u'ПАДЕЖ', u'ЗВАТ'))
elif tag == u'Number=Sing':
stags1.append((u'ЧИСЛО', u'ЕД'))
elif tag == u'Number=Plur':
stags1.append((u'ЧИСЛО', u'МН'))
elif tag == u'Gender=Masc':
stags1.append((u'РОД', u'МУЖ'))
elif tag == u'Gender=Fem':
stags1.append((u'РОД', u'ЖЕН'))
elif tag == u'Gender=Neut':
stags1.append((u'РОД', u'СР'))
else:
print(u'неизвестный тэг "{}"'.format(tag))
raise NotImplementedError()
elif part_of_speech == u'ПРИЛАГАТЕЛЬНОЕ':
if tag == u'Case=Nom':
stags1.append((u'ПАДЕЖ', u'ИМ'))
elif tag == u'Case=Acc':
stags1.append((u'ПАДЕЖ', u'ВИН'))
elif tag == u'Case=Dat':
stags1.append((u'ПАДЕЖ', u'ДАТ'))
elif tag == u'Case=Ins':
stags1.append((u'ПАДЕЖ', u'ТВОР'))
elif tag == u'Case=Prep':
stags1.append((u'ПАДЕЖ', u'ПРЕДЛ'))
elif tag == u'Case=Loc':
stags1.append((u'ПАДЕЖ', u'ПРЕДЛ')) # 03-02-2020 u'МЕСТ'
elif tag == u'Case=Gen':
stags1.append((u'ПАДЕЖ', u'РОД'))
elif tag == u'Number=Sing':
stags1.append((u'ЧИСЛО', u'ЕД'))
elif tag == u'Number=Plur':
stags1.append((u'ЧИСЛО', u'МН'))
elif tag == u'Gender=Masc':
stags1.append((u'РОД', u'МУЖ'))
elif tag == u'Gender=Fem':
stags1.append((u'РОД', u'ЖЕН'))
elif tag == u'Gender=Neut':
stags1.append((u'РОД', u'СР'))
elif tag == u'Degree=Cmp':
stags1.append((u'СТЕПЕНЬ', u'СРАВН'))
elif tag == u'Degree=Pos':
stags1.append((u'СТЕПЕНЬ', u'АТРИБ'))
elif tag in (u'Variant=Short', u'Variant=Brev'):
stags1[short_tag_index] = (u'КРАТКИЙ', u'1')
else:
print(u'неизвестный тэг "{}"'.format(tag))
raise NotImplementedError()
elif part_of_speech == u'ГЛАГОЛ':
if tag == u'Number=Sing':
stags1.append((u'ЧИСЛО', u'ЕД'))
elif tag == u'Number=Plur':
stags1.append((u'ЧИСЛО', u'МН'))
elif tag == u'Gender=Masc':
stags1.append((u'РОД', u'МУЖ'))
elif tag == u'Gender=Fem':
stags1.append((u'РОД', u'ЖЕН'))
elif tag == u'Gender=Neut':
stags1.append((u'РОД', u'СР'))
elif tag == u'Mood=Ind':
stags1.append((u'НАКЛОНЕНИЕ', u'ИЗЪЯВ'))
elif tag == u'Mood=Imp':
stags1.append((u'НАКЛОНЕНИЕ', u'ПОБУД'))
elif tag == u'Tense=Past':
stags1.append((u'ВРЕМЯ', u'ПРОШЕДШЕЕ'))
elif tag == u'Tense=Fut':
stags1.append((u'ВРЕМЯ', u'БУДУЩЕЕ'))
elif tag == u'Tense=Notpast':
stags1.append((u'ВРЕМЯ', u'НАСТОЯЩЕЕ'))
elif tag == u'Tense=Pres':
stags1.append((u'ВРЕМЯ', u'НАСТОЯЩЕЕ'))
elif tag == u'Person=1':
stags1.append((u'ЛИЦО', u'1'))
elif tag == u'Person=2':
stags1.append((u'ЛИЦО', u'2'))
elif tag == u'Person=3':
stags1.append((u'ЛИЦО', u'3'))
elif tag == u'VerbForm=Fin':
pass
elif tag == u'VerbForm=Inf':
pass
elif tag == u'VerbForm=Conv':
pass
else:
msg = u'неизвестный тэг "{}"'.format(tag)
raise RuntimeError(msg)
elif part_of_speech == u'НАРЕЧИЕ':
if tag == u'Degree=Pos':
stags1.append((u'СТЕПЕНЬ', u'АТРИБ'))
elif tag == u'Degree=Cmp':
stags1.append((u'СТЕПЕНЬ', u'СРАВН'))
else:
raise NotImplementedError()
else:
pass
return part_of_speech, stags1
def get_lemma2(self, word, pos_tags):
part_of_speech, decoded_tags = self.decode_pos_tags(pos_tags)
nword = word.lower().replace('ё', 'е')
if nword in self.special_lemmas:
return self.special_lemmas[nword], part_of_speech, decoded_tags
if nword in self.forms:
lemma = self.forms[nword]
return lemma, part_of_speech, decoded_tags
elif nword in self.forms2:
if part_of_speech == 'СУЩЕСТВИТЕЛЬНОЕ':
# Для существительных учитываем падеж.
required_case = None
for tag in decoded_tags:
if tag[0] == 'ПАДЕЖ':
required_case = tag[1]
break
for lemma, lemma_part_of_speech, tag in self.forms2[nword]:
if lemma_part_of_speech == part_of_speech and tag == required_case:
return lemma, part_of_speech, decoded_tags
else:
for lemma, lemma_part_of_speech, tags in self.forms2[nword]:
if lemma_part_of_speech == part_of_speech:
return lemma, part_of_speech, decoded_tags
elif len(word) > 4:
# используем модель лемматизации для OV-слов
ending = nword[-4:]
key = ending + u'|' + part_of_speech
if key in self.key2transducer:
transducer = self.key2transducer[key]
if transducer[0] > 0:
lemma = word[:-transducer[0]] + transducer[1]
else:
lemma = word + transducer[1]
return lemma.lower(), part_of_speech, decoded_tags
# fallback-вариант - возвращаем исходное слово в нижнем регистре в качестве леммы
return nword, part_of_speech, decoded_tags
def lemmatize(self, tagged_words):
"""Для результата работы rupostagger'а добавляем лемму и извлеченный код части речи"""
return [(word, tags,)+tuple(self.get_lemma2(word, tags)) for (word, tags) in tagged_words]
|