Preface
Recording some Python code during the learning process
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
#!/usr/bin/env python
# coding:utf8
import sys
reload(sys)
sys.setdefaultencoding('utf8')
# Load packages
from gensim.models import Word2Vec
from gensim.models.word2vec import LineSentence
# Train model
# sentences = LineSentence('wiki.zh.word.text')
# size: dimension of word vectors
# window: context window size
# min_count: ignore words with frequency below min_count
# model = Word2Vec(sentences, size=128, window=5, min_count=5, workers=4)
# Save model
# model.save('word_embedding_128')
# If the model has already been saved, you can load it directly
# The training and saving code above can all be skipped
model = Word2Vec.load("word_embedding_128")
# Use the model
# Returns multiple most similar words and their corresponding similarity scores
items = model.most_similar(u'中国')
for item in items:
# Word content, word similarity
print item[0], item[1]
# Returns the similarity between two words
model.similarity(u'男人', u'女人')
Tokenization references:
HIT-SCIR LTP Tokenization
jieba Tokenization
Stanford Tokenization