forked from liuhuanyong/SentenceSentimentClassifier
-
Notifications
You must be signed in to change notification settings - Fork 0
/
lstm_sentiment.py
100 lines (86 loc) · 3.9 KB
/
lstm_sentiment.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
#!/usr/bin/env python3
# coding: utf-8
# File: lstm_sentiment.py
# Author: lhy<lhy_in_blcu@126.com,https://huangyong.github.io>
# Date: 18-3-19
import gensim
import numpy as np
from keras.models import load_model
VECTOR_DIR = './embedding/word_vector.bin' # 词向量模型文件
model = gensim.models.KeyedVectors.load_word2vec_format(VECTOR_DIR, binary=False)
'''基于wordvector，通过lookup table的方式找到句子的wordvector的表示'''
def rep_sentencevector(sentence):
word_list = [word for word in sentence.split(' ')]
max_words = 100
embedding_dim = 200
embedding_matrix = np.zeros((max_words, embedding_dim))
for index, word in enumerate(word_list):
try:
embedding_matrix[index] = model[word]
except:
pass
return embedding_matrix
'''构造训练数据'''
def build_traindata():
X_train = list()
Y_train = list()
X_test = list()
Y_test = list()
for line in open('./data/train.txt'):
line = line.strip().strip().split('\t')
sent_vector = rep_sentencevector(line[-1])
X_train.append(sent_vector)
if line[0] == '1':
Y_train.append([0, 1])
else:
Y_train.append([1, 0])
for line in open('./data/test.txt'):
line = line.strip().strip().split('\t')
sent_vector = rep_sentencevector(line[-1])
X_test.append(sent_vector)
if line[0] == '1':
Y_test.append([0, 1])
else:
Y_test.append([1, 0])
return np.array(X_train), np.array(Y_train), np.array(X_test), np.array(Y_test),
'''三层lstm进行训练，迭代20次'''
def train_lstm(X_train, Y_train, X_test, Y_test):
from keras.models import Sequential
from keras.layers import LSTM, Dense
import numpy as np
data_dim = 200 # 对应词向量维度
timesteps = 100 # 对应序列长度
# expected input data shape: (batch_size, timesteps, data_dim)
model = Sequential()
model.add(LSTM(32, return_sequences=True,
input_shape=(timesteps, data_dim))) # returns a sequence of vectors of dimension 32
model.add(LSTM(32, return_sequences=True)) # returns a sequence of vectors of dimension 32
model.add(LSTM(32)) # return a single vector of dimension 32
model.add(Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='rmsprop',
metrics=['accuracy'])
model.fit(X_train, Y_train, batch_size=100, epochs=20, validation_data=(X_test, Y_test))
model.save('./model/sentiment_lstm_model.h5')
'''/
1 [==============================] - 41s 2ms/step - loss: 0.5384 - acc: 0.7142 - val_loss: 0.4223 - val_acc: 0.8281
5 [==============================] - 38s 2ms/step - loss: 0.2885 - acc: 0.8904 - val_loss: 0.3618 - val_acc: 0.8531
10 [==============================] - 40s 2ms/step - loss: 0.1965 - acc: 0.9357 - val_loss: 0.3815 - val_acc: 0.8515
15 [==============================] - 39s 2ms/step - loss: 0.1420 - acc: 0.9577 - val_loss: 0.5172 - val_acc: 0.8501
20 [==============================] - 37s 2ms/step - loss: 0.1055 - acc: 0.9729 - val_loss: 0.5309 - val_acc: 0.8505
'''
'''实际应用，测试'''
def predict_lstm(model_filepath):
model = load_model(model_filepath)
sentence = '这个 电视 真 尼玛 垃圾 ， 老子 再也 不买 了'#[[0.01477097 0.98522896]]
#sentence = '这件 衣服 真的 太 好看 了 ！ 好想 买 啊 '#[[0.9843225 0.01567744]]
sentence_vector = np.array([rep_sentencevector(sentence)])
print(sentence_vector)
print('test after load: ', model.predict(sentence_vector))
if __name__ == '__main__':
# X_train, Y_train, X_test, Y_test = build_traindata()
model_filepath = './model/sentiment_model.h5'
# print(X_train.shape, Y_train.shape)
# print(X_test.shape, Y_test.shape)
# train_lstm(X_train, Y_train, X_test, Y_test)
predict_lstm(model_filepath)