利用卷积神经网络（cnn）实现文本分类

卷积神经网络在情感分析中取得了很好的成果，相比于之前浅层的机器学习方法如NB、SVM效果更好，特别实在数据集较大的情况下，并且CNN不用我们手动去提取特征，原浅层ML是需要进行文本特征提取、文本特征表示、归一化、最后进行文本分类，文本特征提取主要可以分为四步：（1）：对全部训练文档进行分词，由这些词作为向量的维数来表示文本;（2）：统计每一类文档中所有出现的词语及其频率，然后过滤，剔除停用词和单字词;（3）：统计每一类内出现词语的总词频，并取若干个频率更高的词汇作为这一类的特征词集;（4）：去除每一类别中都出现的词，合并所有类别的特征词集，形成总特征词集，最后得到的特征词集是我们用到的特征集合，再用该集合去筛选测试集中的特征。文本的特征表示是利用TF-IDF公式来计算词的权值，这也充分利用的是特征提取时提取的特征来计算特征权值大小的，归一化处理需要处理的数据，经过处理后限制在一定范围内，经过处理后，我们原来的文本信息已经抽象成一个向量化的样本集，然后将样本集和训练好的模板进行相似度计算，若属于该类别，则与其他类别的模板文件进行计算，直到分进相应的类别，这是浅层ML进行文本分类的方式；

CNN进行文本分类相对简单一些，我结合最近做的一些实验总结了一下：

在利用CNN进行文本分类的时候，首先要将原始文本进行预处理，主要还是分词、去除停用词等，然后对预处理后的文本进行向量化利用word2vec，我利用的时word2vec中的skip-gram模型，将搜狗数据集表示为了200维的词向量形式;转化为词向量后就可以将每一句话转化为一个矩阵的形式，这样就跟利用CNN处理图像分类很相似;

说一下实验，我的实验环境：

tensorflow1.2、gpu1050Ti、Ubuntu16.04、pycharm、python2.7

# encoding=utf-8
from __future__ import unicode_literals

import tensorflow as tf
import numpy as np

class TextCNN(object):
"""
使用CNN用于情感分析
整个CNN架构包括词嵌入层，卷积层，max-pooling层和softmax层
"""
def __init__(
self, sequence_length, num_classes,vocab_size,embedding_size, embedding_table,
filter_sizes, num_filters, l2_reg_lambda=0.0):

# 输入，输出，dropout的placeholder
self.input_x = tf.placeholder(tf.int32, [None, sequence_length], name="input_x")
self.input_y = tf.placeholder(tf.float32, [None, num_classes], name="input_y")
self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")

# Keeping track of l2 regularization loss (optional)
l2_loss = tf.constant(0.0)

# 词嵌入层
with tf.device('/cpu:0'), tf.name_scope("embedding"):
W = tf.Variable(embedding_table,name="W")
self.embedded_chars = tf.nn.embedding_lookup(W, self.input_x)
self.embedded_chars_expanded = tf.expand_dims(self.embedded_chars, -1)

# 生成卷积层和max-pooling层
pooled_outputs = []
for i, filter_size in enumerate(filter_sizes):
with tf.name_scope("conv-maxpool-%s" % filter_size):
# Convolution Layer
filter_shape = [filter_size, embedding_size, 1, num_filters]
W = tf.Variable(tf.truncated_normal(filter_shape, stddev=0.1), name="W")
b = tf.Variable(tf.constant(0.1, shape=[num_filters]), name="b")
conv = tf.nn.conv2d(
self.embedded_chars_expanded,
W,
strides=[1, 1, 1, 1],
padding="VALID",
name="conv")
# Apply nonlinearity
# h = tf.nn.relu(tf.nn.bias_add(conv, b), name="relu")
h=tf.nn.relu6(tf.nn.bias_add(conv,b),name="relu")
# Maxpooling over the outputs
# pooled = tf.nn.max_pool(
#     h,
#     ksize=[1, sequence_length - filter_size + 1, 1, 1],
#     strides=[1, 1, 1, 1],
#     padding='VALID',
#     name="pool")
# pooled_outputs.app
1461f
end(pooled)
pooled = tf.nn.avg_pool(
h,
ksize=[1, sequence_length - filter_size + 1, 1, 1],
strides=[1, 1, 1, 1],
padding='VALID',
name="pool")
pooled_outputs.append(pooled)

# 将max-pooling层的各种特征整合在一起
num_filters_total = num_filters * len(filter_sizes)
self.h_pool = tf.concat(pooled_outputs,3)
self.h_pool_flat = tf.reshape(self.h_pool, [-1, num_filters_total])

# 添加全连接层，用于分类
with tf.name_scope("full-connection"):
W_fc1 = tf.Variable(tf.truncated_normal([num_filters_total,500], stddev=0.1))
b_fc1 = tf.Variable(tf.constant(0.1,shape=[500]))
self.h_fc1 = tf.nn.relu6(tf.matmul(self.h_pool_flat, W_fc1) + b_fc1)

# 添加dropout层用于缓和过拟化
with tf.name_scope("dropout"):
# self.h_drop = tf.nn.dropout(self.h_pool_flat, self.dropout_keep_prob)
self.h_drop = tf.nn.dropout(self.h_fc1, self.dropout_keep_prob)

# 产生最后的输出和预测
with tf.name_scope("output"):
# W = tf.get_variable(
#     "W",
#     shape=[num_filters_total, num_classes],
#     initializer=tf.contrib.layers.xavier_initializer())
W = tf.get_variable(
"W",
shape=[500, num_classes],
initializer=tf.contrib.layers.xavier_initializer())
b = tf.Variable(tf.constant(0.1, shape=[num_classes]), name="b")
l2_loss += tf.nn.l2_loss(W)
l2_loss += tf.nn.l2_loss(b)
self.scores = tf.nn.xw_plus_b(self.h_drop, W, b, name="scores")
self.predictions = tf.argmax(self.scores, 1, name="predictions")

# 定义模型的损失函数
with tf.name_scope("loss"):
losses = tf.nn.softmax_cross_entropy_with_logits(logits=self.scores, labels=self.input_y)
self.loss = tf.reduce_mean(losses) + l2_reg_lambda * l2_loss

# 定义模型的准确率
with tf.name_scope("accuracy"):
correct_predictions = tf.equal(self.predictions, tf.argmax(self.input_y, 1))
self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy")

以上时TextCNN的模型结构代码，然后开始进行train，并利用summary和checkpoints来记录模型和训练时的参数等等，利用十折交叉验证来产生准确率，最后利用tensorboard查看accuracy、loss、w、b等等变化图;训练py的代码：
 

#! /usr/bin/env python
# encoding=utf-8
import tensorflow as tf
import numpy as np
import os
import time
import datetime
import data_loader
from cnn_graph import TextCNN
from tensorflow.contrib import learn
from sklearn import cross_validation
import preprocessing
# tf.global_variables
# 伴随tensorflow的summary和checkout
# ==================================================

# Model Hyperparameters
tf.flags.DEFINE_integer("embedding_dim", 200, "Dimensionality of character embedding (default: 128)")
tf.flags.DEFINE_string("filter_sizes", "3,4,5", "Comma-separated filter sizes (default: '3,4,5')")
tf.flags.DEFINE_integer("num_filters", 40, "Number of filters per filter size (default: 128)")
tf.flags.DEFINE_float("dropout_keep_prob", 0.5, "Dropout keep probability (default: 0.5)")
tf.flags.DEFINE_float("l2_reg_lambda", 3.0, "L2 regularizaion lambda (default: 0.0)")

# Training parameters
tf.flags.DEFINE_integer("batch_size", 50, "Batch Size (default: 64)")
tf.flags.DEFINE_integer("num_epochs", 100, "Number of training epochs (default: 200)")
tf.flags.DEFINE_integer("evaluate_every", 100, "Evaluate model on dev set after this many steps (default: 100)")
tf.flags.DEFINE_integer("checkpoint_every", 100, "Save model after this many steps (default: 100)")
# Misc Parameters
tf.flags.DEFINE_boolean("allow_soft_placement", True, "Allow device soft device placement")
tf.flags.DEFINE_boolean("log_device_placement", False, "Log placement of ops on devices")

# w2v文件路径
tf.flags.DEFINE_string("w2v_path", "./w2v_model/retrain_vectors_100.bin", "w2v file")
tf.flags.DEFINE_string("file_dir","./data_process/jd","train/test dataSet")

FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")

# Data Preparatopn
# ==================================================

# Load data
print("Loading data...")
files = ["reviews.neg","reviews.pos"]
# 加载所有的未切分的数据
x_text, y_labels,neg_examples,pos_examples = data_loader.\
load_data_and_labels(data_dir=FLAGS.file_dir,files=files,splitable=False)

# 获取消极数据的2/3,得到的评论的长度离散度更低
neg_accept_length = preprocessing.freq_factor(neg_examples,
percentage=0.8, drawable=False)
neg_accept_length = [item[0] for item in neg_accept_length]
neg_examples = data_loader.load_data_by_length(neg_examples,neg_accept_length)

# 获取积极数据的2/3,得到的评论的长度离散度更低
pos_accept_length = preprocessing.freq_factor(pos_examples,
percentage=0.8, drawable=False)
pos_accept_length = [item[0] for item in pos_accept_length]
pos_examples = data_loader.load_data_by_length(pos_examples,pos_accept_length)

x_text = neg_examples + pos_examples
neg_labels = [[1,0] for _ in neg_examples]
pos_labels = [[0,1] for _ in pos_examples]
y_labels = np.concatenate([neg_labels,pos_labels], axis=0)
print("Loading data finish")

# Build vocabulary
max_document_length = max([len(x.split(" ")) for x in x_text]) # 最长的句子的长度
print(max_document_length)
vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
x = np.array(list(vocab_processor.fit_transform(x_text)))

# 加载提前训练的w2v数据集
word_vecs = data_loader.load_bin_vec(fname=FLAGS.w2v_path,
vocab=list(vocab_processor.vocabulary_._mapping),
ksize=FLAGS.embedding_dim)
# 加载嵌入层的table
W = data_loader.get_W(word_vecs=word_vecs,
vocab_ids_map=vocab_processor.vocabulary_._mapping,
k=FLAGS.embedding_dim,is_rand=False)

# 随机化数据
np.random.seed(10)
shuffle_indices = np.random.permutation(np.arange(len(y_labels)))
x_shuffled = x[shuffle_indices]
y_shuffled = y_labels[shuffle_indices]

out_path = os.path.abspath(os.path.join(os.path.curdir, "runs","parameters"))
parameters = "新全连接+jd数据+10\n" \
"embedding_dim: {},\n" \
"filter_sizes:{},\n" \
"num_filters:{},\n" \
"dropout_keep_prob:{},\n" \
"l2_reg_lambda:{},\n" \
"num_epochs:{},\n" \
"batch_size:{}".format(FLAGS.embedding_dim,FLAGS.filter_sizes,FLAGS.num_filters,
FLAGS.dropout_keep_prob,FLAGS.l2_reg_lambda,FLAGS.num_epochs,
FLAGS.batch_size)
open(out_path, 'w').write(parameters)

# Training
# ==================================================
def train(X_train, X_dev, x_test, y_train, y_dev, y_test):
with tf.Graph().as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess = tf.Session(config=session_conf)
with sess.as_default():
cnn = TextCNN(
sequence_length=max_document_length,
num_classes=2,
vocab_size=len(vocab_processor.vocabulary_),
embedding_size=FLAGS.embedding_dim,
embedding_table=W,
filter_sizes=list(map(int, FLAGS.filter_sizes.split(","))),
num_filters=FLAGS.num_filters,
l2_reg_lambda=FLAGS.l2_reg_lambda)

# Define Training procedure
global_step = tf.Variable(0, name="global_step", trainable=False)
optimizer = tf.train.AdamOptimizer(1e-3)
grads_and_vars = optimizer.compute_gradients(cnn.loss)
train_op = optimizer.apply_gradients(grads_and_vars, global_step=global_step)

# Keep track of gradient values and sparsity (optional)
grad_summaries = []
for g, v in grads_and_vars:
if g is not None:
grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name), g)
sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name), tf.nn.zero_fraction(g))
grad_summaries.append(grad_hist_summary)
grad_summaries.append(sparsity_summary)
grad_summaries_merged = tf.summary.merge(grad_summaries)

# Output directory for models and summaries
timestamp = str(int(time.time()))
out_dir = os.path.abspath(os.path.join(os.path.curdir, "runs", timestamp))
print("Writing to {}\n".format(out_dir))

# Summaries for loss and accuracy
loss_summary = tf.summary.scalar("loss", cnn.loss)
acc_summary = tf.summary.scalar("accuracy", cnn.accuracy)

# Train Summaries
train_summary_op = tf.summary.merge([loss_summary, acc_summary, grad_summaries_merged])
train_summary_dir = os.path.join(out_dir, "summaries", "train")
train_summary_writer = tf.summary.FileWriter(train_summary_dir, sess.graph)

# Dev summaries
dev_summary_op = tf.summary.merge([loss_summary, acc_summary])
dev_summary_dir = os.path.join(out_dir, "summaries", "dev")
dev_summary_writer = tf.summary.FileWriter(dev_summary_dir, sess.graph)

# Checkpoint directory. Tensorflow assumes this directory already exists so we need to create it
checkpoint_dir = os.path.abspath(os.path.join(out_dir, "checkpoints"))
checkpoint_prefix = os.path.join(checkpoint_dir, "model")
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
saver = tf.train.Saver(tf.global_variables())

# Write vocabulary
vocab_processor.save(os.path.join(out_dir, "vocab"))

# Initialize all variables
# sess.run(tf.initialize_all_variables())
sess.run(tf.global_variables_initializer())

def train_step(x_batch, y_batch):
"""
A single training step
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: FLAGS.dropout_keep_prob
}
_, step, summaries, loss, accuracy = sess.run(
[train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
# _, step, loss, accuracy = sess.run(
#     [train_op, global_step, cnn.loss, cnn.accuracy],
#     feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
train_summary_writer.add_summary(summaries, step)

def dev_step(x_batch, y_batch, writer=None):
"""
Evaluates model on a dev set
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: 1.0
}
step, summaries, loss, accuracy = sess.run(
[global_step, dev_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
# step, loss, accuracy = sess.run(
#     [global_step, cnn.loss, cnn.accuracy],
#     feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
if writer:
writer.add_summary(summaries, step)

# Generate batches
batches = data_loader.batch_iter(
list(zip(X_train, y_train)), FLAGS.batch_size, FLAGS.num_epochs)
# Training loop. For each batch...
for batch in batches:
x_batch, y_batch = zip(*batch)
train_step(x_batch, y_batch)
current_step = tf.train.global_step(sess, global_step)
if current_step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
dev_step(X_dev, y_dev, writer=dev_summary_writer)
# dev_step(X_dev, y_dev, writer=None)
print("")
if current_step % FLAGS.checkpoint_every == 0:
path = saver.save(sess, checkpoint_prefix, global_step=current_step)
print("Saved model checkpoint to {}\n".format(path))

# Test loop
# Generate batches for one epoch
batches = data_loader.batch_iter(list(x_test), FLAGS.batch_size, 1, shuffle=False)
# Collect the predictions here
all_predictions = []
for x_test_batch in batches:
batch_predictions = sess.run(cnn.predictions, {cnn.input_x: x_test_batch, cnn.dropout_keep_prob: 1.0})
all_predictions = np.concatenate([all_predictions, batch_predictions])

correct_predictions = float(sum(
all_predictions == np.argmax(y_test,axis=1)))

print("Total number of test examples: {}".format(len(y_test)))
print("Accuracy: {:g}".format(correct_predictions / float(len(y_test))))
# open(os.path.join(out_dir,"test"),'a').write("Accuracy: {:g}".format(correct_predictions / float(len(y_test))))
out_path = os.path.abspath(os.path.join(os.path.curdir, "runs","test"))
open(out_path,'a').write("{:g},".format(correct_predictions / float(len(y_test))))
print("\n写入成功！\n")

# cross-validation
kf = cross_validation.KFold(len(x_shuffled), n_folds=3)
for train_index, test_index in kf:
X_train_total = x_shuffled[train_index]
y_train_total = y_shuffled[train_index]
x_test = x_shuffled[test_index]
y_test = y_shuffled[test_index]

# 分割训练集与验证集
X_train, X_dev, y_train, y_dev = cross_validation.train_test_split(
X_train_total, y_train_total, test_size=0.2, random_state=0)

print("Vocabulary Size: {:d}".format(len(vocab_processor.vocabulary_)))
print("Train/Dev split: {:d}/{:d}".format(len(y_train), len(y_dev)))
训练完成后准确率83%左右，还需要在进一步进行改进来提升正确率，比如利用chunk max-pooling方法代替max-pooling，利用集成的方法，因为word embedding词忽略了当前上下文的含义，潜在认为相同词在不同文本中的含义相同，所以可以利用词义消歧来提升其正确率等等;
训练模型保存在checkpoints中，由model-4000.index,model-4000.meta,model-4000.data等;
最后tensorboard --logdir  /home/yang/PycharmProjects/cnn-text-classification-master/runs/1515468832
/home/yang/PycharmProjects/cnn-text-classification-master/runs/1515468832/checkpoints/model-4300

Total number of test examples: 1333
Accuracy: 0.825956