1. 使用大数据，了解怎么处理数据不能一次全部加载到内存的情况。如果你内存充足，当我没说
  2. 训练好的模型的保存和使用
  3. 使用的模型没变，还是简单的feedforward神经网络（update：添加CNN模型）
  4. 如果你要运行本帖代码，推荐使用GPU版本或强大的VPS，我使用小笔记本差点等吐血
  5. 后续有关于中文的练习《TensorFlow练习13: 制作一个简单的聊天机器人》《TensorFlow练习7: 基于RNN生成古诗词》《TensorFlow练习18: 根据姓名判断性别》

在正文开始之前，我画了一个机器学习模型的基本开发流程图：

TensorFlow练习2: 对评论进行分类

使用的数据集

使用的数据集：http://help.sentiment140.com/for-students/ (情绪分析)

数据集包含1百60万条推特，包含消极、中性和积极tweet。不知道有没有现成的微博数据集。

数据格式：移除表情符号的CSV文件，字段如下：

  • 0 – the polarity of the tweet (0 = negative, 2 = neutral, 4 = positive)
  • 1 – the id of the tweet (2087)
  • 2 – the date of the tweet (Sat May 16 23:58:44 UTC 2009)
  • 3 – the query (lyx). If there is no query, then this value is NO_QUERY.
  • 4 – the user that tweeted (robotickilldozr)
  • 5 – the text of the tweet (Lyx is cool)

training.1600000.processed.noemoticon.csv（238M）
testdata.manual.2009.06.14.csv（74K）

数据预处理

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
import nltk
from nltk . tokenize import word_tokenize
from nltk . stem import WordNetLemmatizer
 
import pickle
import numpy as np
import pandas as pd
from collections import OrderedDict
 
org_train_file = 'training.1600000.processed.noemoticon.csv'
org_test_file = 'testdata.manual.2009.06.14.csv'
 
# 提取文件中有用的字段
def usefull_filed ( org_file , output_file ) :
output = open ( output_file , 'w' )
with open ( org_file , buffering = 10000 , encoding = 'latin-1' ) as f :
try :
for line in f :                  # "4","2193601966","Tue Jun 16 08:40:49 PDT 2009","NO_QUERY","AmandaMarie1028","Just woke up. Having no school is the best feeling ever "
line = line . replace ( '"' , '' )
clf = line . split ( ',' ) [ 0 ]    # 4
if clf == '0' :
clf = [ 0 , 0 , 1 ]    # 消极评论
elif clf == '2' :
clf = [ 0 , 1 , 0 ]    # 中性评论
elif clf == '4' :
clf = [ 1 , 0 , 0 ]    # 积极评论
 
tweet = line . split ( ',' ) [ - 1 ]
outputline = str ( clf ) + ':%:%:%:' + tweet
output . write ( outputline )    # [0, 0, 1]:%:%:%: that's a bummer.  You shoulda got David Carr of Third Day to do it. ;D
except Exception as e :
print ( e )
output . close ( )    # 处理完成，处理后文件大小127.5M
 
usefull_filed ( org_train_file , 'training.csv' )
usefull_filed ( org_test_file , 'tesing.csv' )
 
# 创建词汇表
def create_lexicon ( train_file ) :
lex = [ ]
lemmatizer = WordNetLemmatizer ( )
with open ( train_file , buffering = 10000 , encoding = 'latin-1' ) as f :
try :
count_word = { }    # 统计单词出现次数
for line in f :
tweet = line . split ( ':%:%:%:' ) [ 1 ]
words = word_tokenize ( line . lower ( ) )
for word in words :
word = lemmatizer . lemmatize ( word )
if word not in count_word :
count_word [ word ] = 1
else :
count_word [ word ] += 1
 
count_word = OrderedDict ( sorted ( count_word . items ( ) , key = lambda t : t [ 1 ] ) )
for word in count_word :
if count_word [ word ] < 100000 and count_word [ word ] > 100 :    # 过滤掉一些词
lex . append ( word )
except Exception as e :
print ( e )
return lex
 
lex = create_lexicon ( 'training.csv' )
 
with open ( 'lexcion.pickle' , 'wb' ) as f :
pickle . dump ( lex , f )
 
 
"""
# 把字符串转为向量
def string_to_vector(input_file, output_file, lex):
output_f = open(output_file, 'w')
lemmatizer = WordNetLemmatizer()
with open(input_file, buffering=10000, encoding='latin-1') as f:
for line in f:
label = line.split(':%:%:%:')[0]
tweet = line.split(':%:%:%:')[1]
words = word_tokenize(tweet.lower())
words = [lemmatizer.lemmatize(word) for word in words]
 
features = np.zeros(len(lex))
for word in words:
if word in lex:
features[lex.index(word)] = 1  # 一个句子中某个词可能出现两次,可以用+=1，其实区别不大
features = list(features)
output_f.write(str(label) + ":" + str(features) + '\n')
output_f.close()
 
 
f = open('lexcion.pickle', 'rb')
lex = pickle.load(f)
f.close()
 
# lexcion词汇表大小112k,training.vec大约112k*1600000  170G  太大，只能边转边训练了
# string_to_vector('training.csv', 'training.vec', lex)
# string_to_vector('tesing.csv', 'tesing.vec', lex)
"""

上面代码把原始数据转为training.csv、和tesing.csv，里面只包含label和tweet。lexcion.pickle文件保存了词汇表。

如果数据文件太大，不能一次加载到内存，可以把数据导入数据库
Dask可处理大csv文件

开始漫长的训练

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
import os
import random
import tensorflow as tf
import pickle
import numpy as np
from nltk . tokenize import word _tokenize
from nltk . stem import WordNetLemmatizer
 
f = open ( 'lexcion.pickle' , 'rb' )
lex = pickle . load ( f )
f . close ( )
 
 
def get_random_line ( file , point ) :
file . seek ( point )
file . readline ( )
return file . readline ( )
# 从文件中随机选择n条记录
def get_n_random_line ( file_name , n = 150 ) :
lines = [ ]
file = open ( file_name , encoding = 'latin-1' )
total_bytes = os . stat ( file_name ) . st _size
for i in range ( n ) :
random_point = random . randint ( 0 , total_bytes )
lines . append ( get_random_line ( file , random_point ) )
file . close ( )
return lines
 
 
def get_test_dataset ( test_file ) :
with open ( test_file , encoding = 'latin-1' ) as f :
test_x = [ ]
test_y = [ ]
lemmatizer = WordNetLemmatizer ( )
for line in f :
label = line . split ( ':%:%:%:' ) [ 0 ]
tweet = line . split ( ':%:%:%:' ) [ 1 ]
words = word_tokenize ( tweet . lower ( ) )
words = [ lemmatizer . lemmatize ( word ) for word in words ]
features = np . zeros ( len ( lex ) )
for word in words :
if word in lex :
features [ lex . index ( word ) ] = 1
test_x . append ( list ( features ) )
test_y . append ( eval ( label ) )
return test_x , test _y
 
test_x , test_y = get_test_dataset ( 'tesing.csv' )
 
 
#######################################################################
 
n_input_layer = len ( lex )    # 输入层
 
n_layer_1 = 2000      # hide layer
n_layer_2 = 2000      # hide layer(隐藏层)听着很神秘，其实就是除输入输出层外的中间层
 
n_output_layer = 3        # 输出层
 
 
def neural_network ( data ) :
# 定义第一层"神经元"的权重和biases
layer_1_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_input_layer , n_layer_1 ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_layer_1 ] ) ) }
# 定义第二层"神经元"的权重和biases
layer_2_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_layer_1 , n_layer_2 ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_layer_2 ] ) ) }
# 定义输出层"神经元"的权重和biases
layer_output_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_layer_2 , n_output_layer ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_output_layer ] ) ) }
 
# w·x+b
layer_1 = tf . add ( tf . matmul ( data , layer_1_w_b [ 'w_' ] ) , layer_1_w_b [ 'b_' ] )
layer_1 = tf . nn . relu ( layer_1 )    # 激活函数
layer_2 = tf . add ( tf . matmul ( layer_1 , layer_2_w_b [ 'w_' ] ) , layer_2_w_b [ 'b_' ] )
layer_2 = tf . nn . relu ( layer _2 ) # 激活函数
layer_output = tf . add ( tf . matmul ( layer_2 , layer_output_w_b [ 'w_' ] ) , layer_output_w_b [ 'b_' ] )
 
return layer _output
 
 
X = tf . placeholder ( 'float' )
Y = tf . placeholder ( 'float' )
batch_size = 90
 
def train_neural_network ( X , Y ) :
predict = neural_network ( X )
cost_func = tf . reduce_mean ( tf . nn . softmax_cross_entropy_with_logits ( predict , Y ) )
optimizer = tf . train . AdamOptimizer ( ) . minimize ( cost_func )
 
with tf . Session ( ) as session :
session . run ( tf . initialize_all_variables ( ) )
 
lemmatizer = WordNetLemmatizer ( )
saver = tf . train . Saver ( )
i = 0
pre_accuracy = 0
while True :    # 一直训练
batch_x = [ ]
batch_y = [ ]
 
#if model.ckpt文件已存在:
# saver.restore(session, 'model.ckpt')  恢复保存的session
 
try :
lines = get_n_random_line ( 'training.csv' , batch_size )
for line in lines :
label = line . split ( ':%:%:%:' ) [ 0 ]
tweet = line . split ( ':%:%:%:' ) [ 1 ]
words = word_tokenize ( tweet . lower ( ) )
words = [ lemmatizer . lemmatize ( word ) for word in words ]
 
features = np . zeros ( len ( lex ) )
for word in words :
if word in lex :
features [ lex . index ( word ) ] = 1    # 一个句子中某个词可能出现两次,可以用+=1，其实区别不大
batch_x . append ( list ( features ) )
batch_y . append ( eval ( label ) )
 
session . run ( [ optimizer , cost_func ] , feed_dict = { X : batch_x , Y : batch_y } )
except Exception as e :
print ( e )
 
# 准确率
if i > 100 :
correct = tf . equal ( tf . argmax ( predict , 1 ) , tf . argmax ( Y , 1 ) )
accuracy = tf . reduce_mean ( tf . cast ( correct , 'float' ) )
accuracy = accuracy . eval ( { X : test_x , Y : test_y } )
if accuracy > pre_accuracy :    # 保存准确率最高的训练模型
print ( '准确率: ' , accuracy )
pre_accuracy = accuracy
saver . save ( session , 'model.ckpt' )    # 保存session
i = 0
i += 1
 
 
train_neural_network ( X , Y )

上面程序占用内存600M，峰值1G。

运行：

TensorFlow练习2: 对评论进行分类

训练模型保存为model.ckpt。

使用训练好的模型

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
import tensorflow as tf
import pickle
from nltk . tokenize import word_tokenize
from nltk . stem import WordNetLemmatizer
import numpy as np
 
f = open ( 'lexcion.pickle' , 'rb' )
lex = pickle . load ( f )
f . close ( )
 
n_input_layer = len ( lex )    # 输入层
 
n_layer_1 = 2000      # hide layer
n_layer_2 = 2000      # hide layer(隐藏层)听着很神秘，其实就是除输入输出层外的中间层
 
n_output_layer = 3        # 输出层
def neural_network ( data ) :
# 定义第一层"神经元"的权重和biases
layer_1_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_input_layer , n_layer_1 ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_layer_1 ] ) ) }
# 定义第二层"神经元"的权重和biases
layer_2_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_layer_1 , n_layer_2 ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_layer_2 ] ) ) }
# 定义输出层"神经元"的权重和biases
layer_output_w_b = { 'w_' : tf . Variable ( tf . random_normal ( [ n_layer_2 , n_output_layer ] ) ) , 'b_' : tf . Variable ( tf . random_normal ( [ n_output_layer ] ) ) }
 
# w·x+b
layer_1 = tf . add ( tf . matmul ( data , layer_1_w_b [ 'w_' ] ) , layer_1_w_b [ 'b_' ] )
layer_1 = tf . nn . relu ( layer_1 )    # 激活函数
layer_2 = tf . add ( tf . matmul ( layer_1 , layer_2_w_b [ 'w_' ] ) , layer_2_w_b [ 'b_' ] )
layer_2 = tf . nn . relu ( layer _2 ) # 激活函数
layer_output = tf . add ( tf . matmul ( layer_2 , layer_output_w_b [ 'w_' ] ) , layer_output_w_b [ 'b_' ] )
 
return layer _output
 
X = tf . placeholder ( 'float' )
def prediction ( tweet_text ) :
predict = neural_network ( X )
 
with tf . Session ( ) as session :
session . run ( tf . initialize_all_variables ( ) )
saver = tf . train . Saver ( )
saver . restore ( session , 'model.ckpt' )
 
lemmatizer = WordNetLemmatizer ( )
words = word_tokenize ( tweet_text . lower ( ) )
words = [ lemmatizer . lemmatize ( word ) for word in words ]
 
features = np . zeros ( len ( lex ) )
for word in words :
if word in lex :
features [ lex . index ( word ) ] = 1
#print(predict.eval(feed_dict={X:[features]})) [[val1,val2,val3]]
res = session . run ( tf . argmax ( predict . eval ( feed_dict = { X : [ features ] } ) , 1 ) )
return res
 
 
prediction ( "I am very happe" )

上面使用简单的feedfroward模型，下面使用CNN模型

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
# https://github.com/Lab41/sunny-side-up
import os
import random
import tensorflow as tf
import pickle
import numpy as np
from nltk . tokenize import word_tokenize
from nltk . stem import WordNetLemmatizer
 
f = open ( 'lexcion.pickle' , 'rb' )
lex = pickle . load ( f )
f . close ( )
 
def get_random_line ( file , point ) :
file . seek ( point )
file . readline ( )
return file . readline ( )
# 从文件中随机选择n条记录
def get_n_random_line ( file_name , n = 150 ) :
lines = [ ]
file = open ( file_name , encoding = 'latin-1' )
total_bytes = os . stat ( file_name ) . st_size
for i in range ( n ) :
random_point = random . randint ( 0 , total_bytes )
lines . append ( get_random_line ( file , random_point ) )
file . close ( )
return lines
 
def get_test_dataset ( test_file ) :
with open ( test_file , encoding = 'latin-1' ) as f :
test_x = [ ]
test_y = [ ]
lemmatizer = WordNetLemmatizer ( )
for line in f :
label = line . split ( ':%:%:%:' ) [ 0 ]
tweet = line . split ( ':%:%:%:' ) [ 1 ]
words = word_tokenize ( tweet . lower ( ) )
words = [ lemmatizer . lemmatize ( word ) for word in words ]
features = np . zeros ( len ( lex ) )
for word in words :
if word in lex :
features [ lex . index ( word ) ] = 1
test_x . append ( list ( features ) )
test_y . append ( eval ( label ) )
return test_x , test_y
 
test_x , test_y = get_test_dataset ( 'tesing.csv' )
##############################################################################
input_size = len ( lex )
num_classes = 3
 
X = tf . placeholder ( tf . int32 , [ None , input_size ] )
Y = tf . placeholder ( tf . float32 , [ None , num_classes ] )
 
dropout_keep_prob = tf . placeholder ( tf . float32 )
 
batch_size = 90
 
def neural_network ( ) :
# embedding layer
with tf . device ( '/cpu:0' ) , tf . name_scope ( "embedding" ) :
embedding_size = 128
W = tf . Variable ( tf . random_uniform ( [ input_size , embedding_size ] , - 1.0 , 1.0 ) )
embedded_chars = tf . nn . embedding_lookup ( W , X )
embedded_chars_expanded = tf . expand_dims ( embedded_chars , - 1 )
# convolution + maxpool layer
num_filters = 128
filter_sizes = [ 3 , 4 , 5 ]
pooled_outputs = [ ]
for i , filter_size in enumerate ( filter_sizes ) :
with tf . name_scope ( "conv-maxpool-%s" % filter_size ) :
filter_shape = [ filter_size , embedding_size , 1 , num_filters ]
W = tf . Variable ( tf . truncated_normal ( filter_shape , stddev = 0.1 ) )
b = tf . Variable ( tf . constant ( 0.1 , shape = [ num_filters ] ) )
conv = tf . nn . conv2d ( embedded_chars_expanded , W , strides = [ 1 , 1 , 1 , 1 ] , padding = "VALID" )
h = tf . nn . relu ( tf . nn . bias_add ( conv , b ) )
pooled = tf . nn . max_pool ( h , ksize = [ 1 , input_size - filter_size + 1 , 1 , 1 ] , strides = [ 1 , 1 , 1 , 1 ] , padding = 'VALID' )
pooled_outputs . append ( pooled )
 
num_filters_total = num_filters * len ( filter_sizes )
h_pool = tf . concat ( 3 , pooled_outputs )
h_pool_flat = tf . reshape ( h_pool , [ - 1 , num_filters_total ] )
# dropout
with tf . name_scope ( "dropout" ) :
h_drop = tf . nn . dropout ( h_pool_flat , dropout_keep_prob )
# output
with tf . name_scope ( "output" ) :
W = tf . get_variable ( "W" , shape = [ num_filters_total , num_classes ] , initializer = tf . contrib . layers . xavier_initializer ( ) )
b = tf . Variable ( tf . constant ( 0.1 , shape = [ num_classes ] ) )
output = tf . nn . xw_plus_b ( h_drop , W , b )
return output
 
def train_neural_network ( ) :
output = neural_network ( )
 
optimizer = tf . train . AdamOptimizer ( 1e - 3 )
loss = tf . reduce_mean ( tf . nn . softmax_cross_entropy_with_logits ( output , Y ) )
grads_and_vars = optimizer . compute_gradients ( loss )
train_op = optimizer . apply_gradients ( grads_and_vars )
 
saver = tf . train . Saver ( tf . global_variables ( ) )
with tf . Session ( ) as sess :
sess . run ( tf . global_variables_initializer ( ) )
 
lemmatizer = WordNetLemmatizer ( )
i = 0
while True :
batch_x = [ ]
batch_y = [ ]
 
#if model.ckpt文件已存在:
# saver.restore(session, 'model.ckpt')  恢复保存的session
try :
lines = get_n_random_line ( 'training.csv' , batch_size )
for line in lines :
label = line . split ( ':%:%:%:' ) [ 0 ]
tweet = line . split ( ':%:%:%:' ) [ 1 ]
words = word_tokenize ( tweet . lower ( ) )
words = [ lemmatizer . lemmatize ( word ) for word in words ]
 
features = np . zeros ( len ( lex ) )
for word in words :
if word in lex :
features [ lex . index ( word ) ] = 1    # 一个句子中某个词可能出现两次,可以用+=1，其实区别不大
batch_x . append ( list ( features ) )
batch_y . append ( eval ( label ) )
 
_ , loss_ = sess . run ( [ train_op , loss ] , feed_dict = { X : batch_x , Y : batch_y , dropout_keep_prob : 0.5 } )
print ( loss_ )
except Exception as e :
print ( e )
 
if i % 10 == 0 :
predictions = tf . argmax ( output , 1 )
correct_predictions = tf . equal ( predictions , tf . argmax ( Y , 1 ) )
accuracy = tf . reduce_mean ( tf . cast ( correct_predictions , "float" ) )
accur = sess . run ( accuracy , feed_dict = { X : test_x [ 0 : 50 ] , Y : test_y [ 0 : 50 ] , dropout_keep_prob : 1.0 } )
print ( '准确率:' , accur )
 
i += 1
 
train_neural_network ( )

使用了CNN模型之后，准确率有了显著提升。

Share the post "TensorFlow练习2: 对评论进行分类"

Logo
华为开发者空间

华为开发者空间，是为全球开发者打造的专属开发空间，汇聚了华为优质开发资源及工具，致力于让每一位开发者拥有一台云主机，基于华为根生态开发、创新。

更多推荐

cover

华为云数据仓库专业级开发者认证重磅发布

avatar 华为开发者空间
cover

KubeEdge 新版本5大特性解读：更完备的节点设备能力，全新的Dashboard体验

avatar 华为开发者空间
cover

一文带你搞懂GaussDB数据库性能调优

avatar 华为开发者空间