-
Notifications
You must be signed in to change notification settings - Fork 0
/
word2vec.py
326 lines (295 loc) · 11 KB
/
word2vec.py
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
import pynlpir
from gensim import corpora, models
import gensim
import numpy as np
import scipy.stats as stats
import os
import io
import csv
import json
import math
import sys
import pdb
import chardet
import logging
class LoadData(object):
"""docstring for LoadData"""
def __init__(self):
self.stop_list = self._get_stop_list()
self.docs, self.dictionary, self.corporas = self._part_document()
self.doc_list = self.docs.keys()
def _part_document(self):
docs = {}
for dirname, dirnames,filenames in os.walk('dependence/new_docs'):
for filename in filenames:
path = os.path.join(dirname, filename)
f = open(path)
text = f.readlines()
f.close()
index = filename[:6]
docs[index] = []
for line in text:
for w in line.split(' '):
docs[index].append(w.strip())
dictionary = corpora.Dictionary(docs.values())
corporas = {index: dictionary.doc2bow(docs[index]) for index in docs}
return docs, dictionary, corporas
def _get_stop_list(self,path='dependence/new_list.txt'):
stop_list = []
with io.open(path,'r',encoding='utf-8') as f:
lines = f.readlines()
for l in lines:
l = l.strip()
stop_list.append(l)
return stop_list
def document2sentences(self,document):
pynlpir.open()
words = pynlpir.segment(document,pos_tagging=False)
sign = ['。', ';', '.', ';']
pause_position = []
for i in range(len(words)):
if words[i] in sign: pause_position.append(i)
setences = []
if len(pause_position) == 0:
clean_d = [s.strip() for s in words if s not in self.stop_list]
setences.append(' '.join(clean_d)+'\n')
else:
for i in range(len(pause_position)):
setence = []
if i == 0: setence = words[:pause_position[i]]
elif i == len(pause_position)-1 and i != 0: break
else: setence = words[pause_position[i]:pause_position[i+1]]
clean_s = [s.strip() for s in setence if s not in self.stop_list]
setences.append(' '.join(clean_s)+'\n')
return setences
def write_sentences_per_doc(self):
for dirname, dirnames,filenames in os.walk('dependence/new_data'):
for filename in filenames:
read_path = os.path.join(dirname, filename)
rf = open(read_path, 'r',encoding='utf-8')
text = rf.readlines()
rf.close()
document = ''.join(text).replace('\n','')
if document == '': print(filename)
setences = self.document2sentences(document)
if not setences: print(filename)
write_path = os.path.join('dependence/new_docs', filename)
wf = open(write_path, 'w')
wf.writelines(setences)
wf.close()
def line_ss():
lines = []
for dirname, dirnames,filenames in os.walk('dependence/new_docs'):
for filename in filenames:
read_path = os.path.join(dirname, filename)
rf = open(read_path, 'r',encoding='utf-8')
# pdb.set_trace()
document = rf.readlines()
lines.extend(document)
rf.close()
wf = open('dependence/input_sentences.txt', 'w')
wf.writelines(lines)
wf.close()
class Word2Vec(object):
"""docstring for Word2Vec"""
def __init__(self,docs,dictionary,corporas,word2vec_path,size=300):
self.size = size
self.docs = docs
self.dictionary = dictionary
self.corporas = corporas
self.tfidf = self.__tfidf_model()
self.model = self.load_word2vec_model(word2vec_path)
self.doc_vectors = self.get_doc_vector_map()
self.key_docs_map = self.get_key_docs_map()
def __tfidf_model(self):
tfidf = gensim.models.tfidfmodel.TfidfModel(self.corporas.values())
return tfidf
def load_word2vec_model(self,path):
model = gensim.models.word2vec.Word2Vec.load(path)
return model
def get_doc_tfidf(self,index):
doc_tfidf = self.tfidf[self.corporas[index]]
return doc_tfidf
def get_doc_keys(self,index,top=10,val=0.1):
doc_tfidf = self.get_doc_tfidf(index)
doc_tfidf.sort(key=lambda x: x[1])
keys = [k for k in doc_tfidf if k[0]>val]
return keys
def get_key_docs_map(self):
key_docs_map = {}
doc_keys_map = {d: self.get_doc_keys(d) for d in self.docs}
for doc,keys in doc_keys_map.items():
for k in keys:
if k[0] not in key_docs_map: key_docs_map[k[0]] = [(doc,k[1])]
else: key_docs_map[k[0]].append((doc,k[1]))
for key,doc in key_docs_map.items():
doc.sort(key=lambda x: x[1])
key_docs_map[key] = set([d[0] for d in doc])
return key_docs_map
def get_word_vector(self,wordid):
word = self.dictionary[wordid]
word_vector = self.model[word]
return self.model[word]
def get_doc_vector(self,index):
doc_tfidf = self.get_doc_tfidf(index)
doc_vector = np.array([0]*self.size,dtype=float)
word_count = 0
for w in doc_tfidf:
wordid,tfidf = w[:]
if self.dictionary[wordid] in self.model.vocab.keys():
word_count += 1
word_vector = self.get_word_vector(wordid)
doc_vector += tfidf*np.array(word_vector)/word_count
else: print(self.dictionary[wordid])
return doc_vector
def get_doc_vector_map(self):
vector_map = {}
for d in self.docs:
vector_map[d] = self.get_doc_vector(d)
return vector_map
class Search(object):
"""docstring for Query"""
def __init__(self,model_path='dependence/word2vec/word2vec_size_300'):
data = LoadData()
self.stop_list = set(data.stop_list)
self.w2v = Word2Vec(data.docs,data.dictionary,data.corporas,model_path,size=300)
self.vocab = self.w2v.model.vocab
self.word2id = {word:id for id,word in self.w2v.dictionary.iteritems()}
def query2words(self,query):
words = []
segs = query.split(' ')
for s in segs:
s = s.strip() ## need regularization
if s in self.vocab: words.append(s) ## in word2vec vocab
else:
pynlpir.open()
# words.extend(pynlpir.get_key_words(query,max_words=3))
word_segs = pynlpir.segment(query,pos_tagging=False)
for word in word_segs:
if word not in self.stop_list: words.append(word)
print(words)
return words
def query2vector(self,words):
query_vector = np.array([0]*self.w2v.size,dtype=float)
# words = self.query2words(query)
## if all words in dictionary, we can use tfidf
# dictionary = self.w2v.dictionary
# query_vector = np.array([0]*self.w2v.size,dtype=float)
# query_corpora = dictionary.doc2bow(words)
# query_tfidf = self.w2v.tfidf[query_corpora]
# for w in query_tfidf:
# wordid,tfidf = w[:]
# word_vector = self.w2v.get_word_vector(wordid)
# query_vector += tfidf*np.array(word_vector)
n = len(words)
if n>0:
for word in words:
query_vector += (1/n)*self.w2v.model[word]
return query_vector
def query_key_match(self,words):
key_docs = self.w2v.key_docs_map
doc_set = set()
for word in words:
if word in self.word2id:
wordid = self.word2id[word]
if wordid in key_docs:
doc_set.update(key_docs[wordid])
else: print(str(word)+' is not in dictionary!')
return doc_set
def query_vector_relative_docs(self,query,top=100,val=0.3):
words = self.query2words(query)
key_match_docs = self.query_key_match(words)
query_vector = self.query2vector(words)
docs = []; return_list = []
for doc,vec in self.w2v.doc_vectors.items():
sim = cosin_simiarity(query_vector,vec)
if sim>val: docs.append([sim,doc])
if docs:
docs.sort(reverse=True)
if key_match_docs:
return_list = [d[1] for d in docs if d[1] in key_match_docs]
else: return_list = [d[1] for d in docs] # if return_list is null, return query_key_match
return return_list
def word_doc_sim(self,word,doc):
sim = 0
if word in self.vocab:
word_vector = self.w2v.model[word]
sim = cosin_simiarity(self.w2v.doc_vectors[doc],word_vector)
return sim
def query_words_relative_docs(self,query,top=100,val=0.3):
words = self.query2words(query)
docs = []; return_list = []
if words:
for doc in self.w2v.doc_vectors:
avg_sim = np.average([self.word_doc_sim(w,doc) for w in words])
if avg_sim>val: docs.append([avg_sim,doc])
if docs:
docs.sort(reverse=True)
return_list = [d[1] for d in docs]
return return_list
def word_for_docs(self,word,val):
doc_sim = {}
if word in self.vocab:
query_vector = self.w2v.model[word]
for doc,vec in self.w2v.doc_vectors.items():
sim = cosin_simiarity(vec,query_vector)
if sim>val: doc_sim[doc] = sim
if not doc_sim: print('no documents is relative to the word with correlation of '+str(val))
else: print(word + 'not in vocab')
return doc_sim
search = Search()
def doc_search(query,val=0.3):
return search.query_vector_relative_docs(query,val)
def cosin_simiarity(vector_a,vector_b):
norm_a = np.linalg.norm(vector_a)
norm_b = np.linalg.norm(vector_b)
inner_ab = np.dot(vector_a,vector_b)
return inner_ab/(norm_a*norm_b)
def save_document_similarity_matrix(model,document_list,similarity=cosin_simiarity,path='dependence/similarity/'):
similarity_matrix = {}
for c in document_list:
similarity_matrix[c] = {}
for d in document_list:
if c != d:
if d in similarity_matrix: similarity_matrix[c][d] = similarity_matrix[d][c]
else:
vector_c = model.get_doc_vector(c)
vector_d = model.get_doc_vector(d)
similarity_matrix[c][d] = str(similarity(vector_c,vector_d))
print((c,d))
with open(path+c+'.json','w') as f: json.dump(similarity_matrix[c],f)
return similarity_matrix
def train_w2v_model(size,save_path='dependence/word2vec/word2vec_2'):
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
sentences = gensim.models.word2vec.LineSentence('dependence/corporas/wiki_cn.txt')
model = gensim.models.word2vec.Word2Vec(size=size,window=5,min_count=5,workers=4)
model.build_vocab(sentences)
model.train(sentences)
model.save(save_path)
return model
def test():
l = LoadData()
w2v = Word2Vec(l.docs,l.dictionary,l.corporas,size=300)
save_document_similarity_matrix(w2v,l.doc_list,path='dependence/word2vec_similarity/')
def part_sentence(stop_list):
pynlpir.open()
for dirname, dirnames,filenames in os.walk('dependence/ch_corporas/wiki/lost'):
for filename in filenames:
lines = []
read_path = os.path.join(dirname, filename)
rf = open(read_path,'rb')
print(filename)
for line in rf:
# detector.feed(byte)
encoding = chardet.detect(line)['encoding']
if encoding == None: encoding = 'utf-8'
new_line = line.decode(encoding,'ignore')
words = pynlpir.segment(new_line,pos_tagging=False)
clean_words = [w.strip() for w in words if w not in stop_list]
str_line = ' '.join(clean_words)
if str_line: lines.append(str_line+'\n')
rf.close()
write_path = os.path.join('dependence/ch_corporas/wiki_clean', filename)
wf = open(write_path, 'w')
wf.writelines(lines)
wf.close()