Tensorflow深度學習使用CNN分類英文文本

前言

Github源碼地址

本文同時也是學習唐宇迪老師深度學習課程的一些理解與記錄。

文中代碼是實現在TensorFlow下使用卷積神經網絡(CNN)做英文文本的分類任務(本次是垃圾郵件的二分類任務),當然垃圾郵件分類是一種應用環境,模型方法也可以推廣到其它應用場景,如電商商品好評差評分類、正負面新聞等。

這裡寫圖片描述

源碼與數據

源碼

– data_helpers.py

– train.py

– text_cnn.py

– eval.py(Save the evaluations to a csv, in case the user wants to inspect,analyze, or otherwise use the classifications generated by the neural net)

數據

– rt-polarity.neg

– rt-polarity.pos

這裡寫圖片描述

這裡寫圖片描述

train.py 源碼及分析

import tensorflow as tf
import numpy as np
import os
import time
import datetime
import data_helpers
from text_cnn import TextCNN
from tensorflow.contrib import learn
# Parameters
# ==================================================
# Data loading params
# 語料文件路徑定義
tf.flags.DEFINE_float("dev_sample_percentage", .1, "Percentage of the training data to use for validation")
tf.flags.DEFINE_string("positive_data_file", "./data/rt-polaritydata/rt-polarity.pos", "Data source for the positive data.")
tf.flags.DEFINE_string("negative_data_file", "./data/rt-polaritydata/rt-polarity.neg", "Data source for the negative data.")

# Model Hyperparameters
# 定義網絡超參數
tf.flags.DEFINE_integer("embedding_dim", 128, "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", 128, "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", 0.0, "L2 regularization lambda (default: 0.0)")

# Training parameters
# 訓練參數
tf.flags.DEFINE_integer("batch_size", 32, "Batch Size (default: 32)")
tf.flags.DEFINE_integer("num_epochs", 200, "Number of training epochs (default: 200)") # 總訓練次數
tf.flags.DEFINE_integer("evaluate_every", 100, "Evaluate model on dev set after this many steps (default: 100)") # 每訓練100次測試一下
tf.flags.DEFINE_integer("checkpoint_every", 100, "Save model after this many steps (default: 100)") # 保存一次模型
tf.flags.DEFINE_integer("num_checkpoints", 5, "Number of checkpoints to store (default: 5)")
# Misc Parameters
tf.flags.DEFINE_boolean("allow_soft_placement", True, "Allow device soft device placement") # 加上一個佈爾類型的參數,要不要自動分配
tf.flags.DEFINE_boolean("log_device_placement", False, "Log placement of ops on devices") # 加上一個佈爾類型的參數,要不要打印日志

# 打印一下相關初始參數
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
    print("{}={}".format(attr.upper(), value))
print("")

# Data Preparation
# ==================================================
# Load data
print("Loading data...")
x_text, y = data_helpers.load_data_and_labels(FLAGS.positive_data_file, FLAGS.negative_data_file)
# Build vocabulary
max_document_length = max([len(x.split(" ")) for x in x_text]) # 計算最長郵件
vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length) # tensorflow提供的工具,將數據填充為最大長度,默認0填充
x = np.array(list(vocab_processor.fit_transform(x_text)))

# Randomly shuffle data
# 數據洗牌
np.random.seed(10)
# np.arange生成隨機序列
shuffle_indices = np.random.permutation(np.arange(len(y)))
x_shuffled = x[shuffle_indices]
y_shuffled = y[shuffle_indices]

# 將數據按訓練train和測試dev分塊
# Split train/test set
# TODO: This is very crude, should use cross-validation
dev_sample_index = -1 * int(FLAGS.dev_sample_percentage * float(len(y)))
x_train, x_dev = x_shuffled[:dev_sample_index], x_shuffled[dev_sample_index:]
y_train, y_dev = y_shuffled[:dev_sample_index], y_shuffled[dev_sample_index:]
print("Vocabulary Size: {:d}".format(len(vocab_processor.vocabulary_)))
print("Train/Dev split: {:d}/{:d}".format(len(y_train), len(y_dev))) # 打印切分的比例
# Training
# ==================================================
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():
        # 卷積池化網絡導入
        cnn = TextCNN(
            sequence_length=x_train.shape[1],
            num_classes=y_train.shape[1], # 分幾類
            vocab_size=len(vocab_processor.vocabulary_),
            embedding_size=FLAGS.embedding_dim,
            filter_sizes=list(map(int, FLAGS.filter_sizes.split(","))), # 上面定義的filter_sizes拿過來,"3,4,5"按","分割
            num_filters=FLAGS.num_filters, # 一共有幾個filter
            l2_reg_lambda=FLAGS.l2_reg_lambda) # l2正則化項

        # Define Training procedure
        global_step = tf.Variable(0, name="global_step", trainable=False)
        optimizer = tf.train.AdamOptimizer(1e-3) # 定義優化器
        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
        # 損失函數和準確率的參數保存
        loss_summary = tf.summary.scalar("loss", cnn.loss)
        acc_summary = tf.summary.scalar("accuracy", cnn.accuracy)

        # Train Summaries
        # 訓練數據保存
        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
        # 測試數據保存
        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(), max_to_keep=FLAGS.num_checkpoints) # 前面定義好參數num_checkpoints

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

        # Initialize all variables
        sess.run(tf.global_variables_initializer()) # 初始化所有變量

        # 定義訓練函數
        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 # 參數在前面有定義
            }
            _, step, summaries, loss, accuracy = sess.run(
                [train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy], feed_dict)
            time_str = datetime.datetime.now().isoformat() # 取當前時間,python的函數
            print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
            train_summary_writer.add_summary(summaries, step)

        # 定義測試函數
        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 # 神經元全部保留
            }
            step, summaries, loss, accuracy = sess.run(
                [global_step, dev_summary_op, 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_helpers.batch_iter(list(zip(x_train, y_train)), FLAGS.batch_size, FLAGS.num_epochs)
        # Training loop. For each batch...
        # 訓練部分
        for batch in batches:
            x_batch, y_batch = zip(*batch) # 按batch把數據拿進來
            train_step(x_batch, y_batch)
            current_step = tf.train.global_step(sess, global_step) # 將Session和global_step值傳進來
            if current_step % FLAGS.evaluate_every == 0: # 每FLAGS.evaluate_every次每100執行一次測試
                print("\nEvaluation:")
                dev_step(x_dev, y_dev, writer=dev_summary_writer)
                print("")
            if current_step % FLAGS.checkpoint_every == 0: # 每checkpoint_every次執行一次保存模型
                path = saver.save(sess, './', global_step=current_step) # 定義模型保存路徑
                print("Saved model checkpoint to {}\n".format(path))

data_helpers.py 源碼及分析

import numpy as np
import re
import itertools
from collections import Counter

def clean_str(string):
    """
    Tokenization/string cleaning for all datasets except for SST.
    Original taken from https://github.com/yoonkim/CNN_sentence/blob/master/process_data.py
    """
    # 清理數據替換掉無詞義的符號
    string = re.sub(r"[^A-Za-z0-9(),!?\'\`]", " ", string)
    string = re.sub(r"\'s", " \'s", string)
    string = re.sub(r"\'ve", " \'ve", string)
    string = re.sub(r"n\'t", " n\'t", string)
    string = re.sub(r"\'re", " \'re", string)
    string = re.sub(r"\'d", " \'d", string)
    string = re.sub(r"\'ll", " \'ll", string)
    string = re.sub(r",", " , ", string)
    string = re.sub(r"!", " ! ", string)
    string = re.sub(r"\(", " \( ", string)
    string = re.sub(r"\)", " \) ", string)
    string = re.sub(r"\?", " \? ", string)
    string = re.sub(r"\s{2,}", " ", string)
    return string.strip().lower()
def load_data_and_labels(positive_data_file, negative_data_file):
    """
    Loads MR polarity data from files, splits the data into words and generates labels.
    Returns split sentences and labels.
    """
    # Load data from files
    positive = open(positive_data_file, "rb").read().decode('utf-8')
    negative = open(negative_data_file, "rb").read().decode('utf-8')

    # 按回車分割樣本
    positive_examples = positive.split('\n')[:-1]
    negative_examples = negative.split('\n')[:-1]

    # 去空格
    positive_examples = [s.strip() for s in positive_examples]
    negative_examples = [s.strip() for s in negative_examples]

    #positive_examples = list(open(positive_data_file, "rb").read().decode('utf-8'))
    #positive_examples = [s.strip() for s in positive_examples]
    #negative_examples = list(open(negative_data_file, "rb").read().decode('utf-8'))
    #negative_examples = [s.strip() for s in negative_examples]
    # Split by words
    x_text = positive_examples + negative_examples
    x_text = [clean_str(sent) for sent in x_text] # 字符過濾,實現函數見clean_str()
    # Generate labels
    positive_labels = [[0, 1] for _ in positive_examples]
    negative_labels = [[1, 0] for _ in negative_examples]
    y = np.concatenate([positive_labels, negative_labels], 0) # 將兩種label連在一起
    return [x_text, y]

# 創建batch迭代模塊
def batch_iter(data, batch_size, num_epochs, shuffle=True): # shuffle=True洗牌
    """
    Generates a batch iterator for a dataset.
    """
    # 每次隻輸出shuffled_data[start_index:end_index]這麼多
    data = np.array(data)
    data_size = len(data)
    num_batches_per_epoch = int((len(data)-1)/batch_size) + 1 # 每一個epoch有多少個batch_size
    for epoch in range(num_epochs):
        # Shuffle the data at each epoch
        if shuffle:
            shuffle_indices = np.random.permutation(np.arange(data_size)) # 洗牌
            shuffled_data = data[shuffle_indices]
        else:
            shuffled_data = data
        for batch_num in range(num_batches_per_epoch):
            start_index = batch_num * batch_size # 當前batch的索引開始
            end_index = min((batch_num + 1) * batch_size, data_size) # 判斷下一個batch是不是超過最後一個數據瞭
            yield shuffled_data[start_index:end_index]

text_cnn.py 源碼及分析

import tensorflow as tf
import numpy as np
# 定義CNN網絡實現的類
class TextCNN(object):
    """
    A CNN for text classification.
    Uses an embedding layer, followed by a convolutional, max-pooling and softmax layer.
    """
    def __init__(self, sequence_length, num_classes, vocab_size,
                 embedding_size, filter_sizes, num_filters, l2_reg_lambda=0.0): # 把train.py中TextCNN裡定義的參數傳進來

        # Placeholders for input, output and dropout
        self.input_x = tf.placeholder(tf.int32, [None, sequence_length], name="input_x") # input_x輸入語料,待訓練的內容,維度是sequence_length,"N個詞構成的N維向量"
        self.input_y = tf.placeholder(tf.float32, [None, num_classes], name="input_y") # input_y輸入語料,待訓練的內容標簽,維度是num_classes,"正面 || 負面"
        self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob") # dropout_keep_prob dropout參數,防止過擬合,訓練時用
        # Keeping track of l2 regularization loss (optional)
        l2_loss = tf.constant(0.0) # 先不用,寫0
        # Embedding layer
        # 指定運算結構的運行位置在cpu非gpu,因為"embedding"無法運行在gpu
        # 通過tf.name_scope指定"embedding"
        with tf.device('/cpu:0'), tf.name_scope("embedding"): # 指定cpu
            self.W = tf.Variable(tf.random_uniform([vocab_size, embedding_size], -1.0, 1.0), name="W") # 定義W並初始化
            self.embedded_chars = tf.nn.embedding_lookup(self.W, self.input_x)
            self.embedded_chars_expanded = tf.expand_dims(self.embedded_chars, -1) # 加一個維度,轉換為4維的格式
        # Create a convolution + maxpool layer for each filter size
        pooled_outputs = []
        # filter_sizes卷積核尺寸,枚舉後遍歷
        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] # 4個參數分別為filter_size高h,embedding_size寬w,channel為1,filter個數
                W = tf.Variable(tf.truncated_normal(filter_shape, stddev=0.1), name="W") # W進行高斯初始化
                b = tf.Variable(tf.constant(0.1, shape=[num_filters]), name="b") # b給初始化為一個常量
                conv = tf.nn.conv2d(
                    self.embedded_chars_expanded,
                    W,
                    strides=[1, 1, 1, 1],
                    padding="VALID", # 這裡不需要padding
                    name="conv")
                # Apply nonlinearity 激活函數
                # 可以理解為,正面或者負面評價有一些標志詞匯,這些詞匯概率被增強,即一旦出現這些詞匯,傾向性分類進正或負面評價,
                # 該激勵函數可加快學習進度,增加稀疏性,因為讓確定的事情更確定,噪聲的影響就降到瞭最低。
                h = tf.nn.relu(tf.nn.bias_add(conv, b), name="relu")
                # Maxpooling over the outputs
                # 池化
                pooled = tf.nn.max_pool(
                    h,
                    ksize=[1, sequence_length - filter_size + 1, 1, 1], # (h-filter+2padding)/strides+1=h-f+1
                    strides=[1, 1, 1, 1],
                    padding='VALID', # 這裡不需要padding
                    name="pool")
                pooled_outputs.append(pooled)

        # Combine all the pooled features
        num_filters_total = num_filters * len(filter_sizes)
        self.h_pool = tf.concat(3, pooled_outputs)
        self.h_pool_flat = tf.reshape(self.h_pool, [-1, num_filters_total]) # 扁平化數據,跟全連接層相連
        # Add dropout
        # drop層,防止過擬合,參數為dropout_keep_prob
        # 過擬合的本質是采樣失真,噪聲權重影響瞭判斷,如果采樣足夠多,足夠充分,噪聲的影響可以被量化到趨近事實,也就無從過擬合。
        # 即數據越大,drop和正則化就越不需要。
        with tf.name_scope("dropout"):
            self.h_drop = tf.nn.dropout(self.h_pool_flat, self.dropout_keep_prob)

        # Final (unnormalized) scores and predictions
        # 輸出層
        with tf.name_scope("output"):
            W = tf.get_variable(
                "W",
                shape=[num_filters_total, num_classes], #前面連扁平化後的池化操作
                initializer=tf.contrib.layers.xavier_initializer()) # 定義初始化方式
            b = tf.Variable(tf.constant(0.1, shape=[num_classes]), name="b")
            # 損失函數導入
            l2_loss += tf.nn.l2_loss(W)
            l2_loss += tf.nn.l2_loss(b)
            # xw+b
            self.scores = tf.nn.xw_plus_b(self.h_drop, W, b, name="scores") # 得分函數
            self.predictions = tf.argmax(self.scores, 1, name="predictions") # 預測結果

        # CalculateMean cross-entropy loss
        with tf.name_scope("loss"):
            # loss,交叉熵損失函數
            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

        # Accuracy
        with tf.name_scope("accuracy"):
            # 準確率,求和計算算數平均值
            correct_predictions = tf.equal(self.predictions, tf.argmax(self.input_y, 1))
            self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy")

這裡寫圖片描述

以上就是Tensorflow深度學習CNN實現英文文本分類的詳細內容,更多關於Tensorflow實現CNN分類英文文本的資料請關註WalkonNet其它相關文章!

推薦閱讀: