NLP文本分类实战（四）--word2vec的skip-gram实现

上一篇博客，详细讲解了word2vec的两个模型：CBOW和skip-gram

Word2Vec之Skip-Gram模型、

下面代码将用TensorFlow实现Word2Vec中的Skip-Gram模型。

1、导入相关包

mport time
import numpy as np
import tensorflow as tf
import random
from collections import Counter

2、载入相关的数据文件

使用的数据是来自维基百科文章，数据集. 该数据集已经去除了特殊的符号。

with open('data/text8') as f:
    text = f.read()

3、数据预处理

数据预处理包括：

1）替换文本中的特殊符号并去除低频词（或者说选出最长出现的前N个词）；

2）将分本切分为多个单词

3）构建语料；

4）单词映射（构建索引和单词的字典）

# 定义函数来完成数据的预处理
def preprocess(text, freq=5):
    '''
    对文本进行预处理
    
    参数
    ---
    text: 文本数据
    freq: 词频阈值
    '''
    # 对文本中的符号进行替换
    text = text.lower()
    text = text.replace('.', ' <PERIOD> ')
    text = text.replace(',', ' <COMMA> ')
    text = text.replace('"', ' <QUOTATION_MARK> ')
    text = text.replace(';', ' <SEMICOLON> ')
    text = text.replace('!', ' <EXCLAMATION_MARK> ')
    text = text.replace('?', ' <QUESTION_MARK> ')
    text = text.replace('(', ' <LEFT_PAREN> ')
    text = text.replace(')', ' <RIGHT_PAREN> ')
    text = text.replace('--', ' <HYPHENS> ')
    text = text.replace('?', ' <QUESTION_MARK> ')
    # text = text.replace('\n', ' <NEW_LINE> ')
    text = text.replace(':', ' <COLON> ')
    words = text.split()
    
    # 删除低频词，减少噪音影响
    word_counts = Counter(words)
    trimmed_words = [word for word in words if word_counts[word] > freq]

    return trimmed_words

清洗文本、分词，并构建映射表

#清洗文本并分词
words = preprocess(text)
print(words[:20])

# 构建映射表
vocab = set(words)
vocab_to_int = {w: c for c, w in enumerate(vocab)}
int_to_vocab = {c: w for c, w in enumerate(vocab)}


print("total words: {}".format(len(words)))
print("unique words: {}".format(len(set(words))))


# 对原文本进行vocab到int的转换
int_words = [vocab_to_int[w] for w in words]

4、采样

对停用词进行采样，例如“the”， “of”以及“for”这类单词进行剔除。剔除这些单词以后能够加快我们的训练过程，同时减少训练过程中的噪音。详细可以看https://github.com/DongYongfei36/zhihu/blob/master/skip_gram/Skip-Gram-English-Corpus.ipynb

t = 1e-5 # t值
threshold = 0.8 # 剔除概率阈值

# 统计单词出现频次
int_word_counts = Counter(int_words)
total_count = len(int_words)
# 计算单词频率
word_freqs = {w: c/total_count for w, c in int_word_counts.items()}
# 计算被删除的概率
prob_drop = {w: 1 - np.sqrt(t / word_freqs[w]) for w in int_word_counts}
# 对单词进行采样
train_words = [w for w in int_words if prob_drop[w] < threshold]

print（len(train_words))

5、定义一个函数，用于生成skip-gram模型用的batch

对于一个给定词，离它越近的词可能与它越相关，离它越远的词越不相关，这里我们设置窗口大小为5，对于每个训练单词，我们还会在[1:5]之间随机生成一个整数R，用R作为我们最终选择output word的窗口大小。这里之所以多加了一步随机数的窗口重新选择步骤，是为了能够让模型更聚焦于当前input word的邻近词。

def get_targets(words, idx, window_size=5):
    '''
    获得input word的上下文单词列表
    
    参数
    ---
    words: 单词列表
    idx: input word的索引号
    window_size: 窗口大小
    '''
    target_window = np.random.randint(1, window_size+1)
    # 这里要考虑input word前面单词不够的情况
    start_point = idx - target_window if (idx - target_window) > 0 else 0
    end_point = idx + target_window
    # output words(即窗口中的上下文单词)
    targets = set(words[start_point: idx] + words[idx+1: end_point+1])
    return list(targets)

def get_batches(words, batch_size, window_size=5):
    '''
    构造一个获取batch的生成器
    '''
    n_batches = len(words) // batch_size
    
    # 仅取full batches
    words = words[:n_batches*batch_size]
    
    for idx in range(0, len(words), batch_size):
        x, y = [], []
        batch = words[idx: idx+batch_size]
        for i in range(len(batch)):
            batch_x = batch[i]
            batch_y = get_targets(batch, i, window_size)
            # 由于一个input word会对应多个output word，因此需要长度统一
            x.extend([batch_x]*len(batch_y))
            y.extend(batch_y)
        yield x, y

6、建立模型

train_graph = tf.Graph()
with train_graph.as_default():
    inputs = tf.placeholder(tf.int32, shape=[None], name='inputs')
    labels = tf.placeholder(tf.int32, shape=[None, None], name='labels')

vocab_size = len(int_to_vocab)
embedding_size = 200 # 嵌入维度

with train_graph.as_default():
    # 嵌入层权重矩阵
    embedding = tf.Variable(tf.random_uniform([vocab_size, embedding_size], -1, 1))
    # 实现lookup
    embed = tf.nn.embedding_lookup(embedding, inputs)

n_sampled = 100

with train_graph.as_default():
    softmax_w = tf.Variable(tf.truncated_normal([vocab_size, embedding_size], stddev=0.1))
    softmax_b = tf.Variable(tf.zeros(vocab_size))
    # 计算negative sampling下的损失
    loss = tf.nn.sampled_softmax_loss(softmax_w, softmax_b, labels, embed, n_sampled, vocab_size)
    cost = tf.reduce_mean(loss)
    optimizer = tf.train.AdamOptimizer().minimize(cost)

with train_graph.as_default():
    # 随机挑选一些单词
    valid_size = 16 
    valid_window = 100
    # 从不同位置各选8个单词
    valid_examples = np.array(random.sample(range(valid_window), valid_size//2))
    valid_examples = np.append(valid_examples, 
                               random.sample(range(1000,1000+valid_window), valid_size//2))
    
    valid_size = len(valid_examples)
    # 验证单词集
    valid_dataset = tf.constant(valid_examples, dtype=tf.int32)
    
    # 计算每个词向量的模并进行单位化
    norm = tf.sqrt(tf.reduce_sum(tf.square(embedding), 1, keep_dims=True))
    normalized_embedding = embedding / norm
    # 查找验证单词的词向量
    valid_embedding = tf.nn.embedding_lookup(normalized_embedding, valid_dataset)
    # 计算余弦相似度
    similarity = tf.matmul(valid_embedding, tf.transpose(normalized_embedding))

7、开始训练

epochs = 10 # 迭代轮数
batch_size = 64#1000 # batch大小
window_size = 10 # 窗口大小

with train_graph.as_default():
    saver = tf.train.Saver() # 文件存储

with tf.Session(graph=train_graph) as sess:
    iteration = 1
    loss = 0
    sess.run(tf.global_variables_initializer())

    for e in range(1, epochs+1):
        batches = get_batches(train_words, batch_size, window_size)
        start = time.time()
        # 
        for x, y in batches:
            
            feed = {inputs: x,
                    labels: np.array(y)[:, None]}
            train_loss, _ = sess.run([cost, optimizer], feed_dict=feed)
            
            loss += train_loss
            
            if iteration % 100 == 0: 
                end = time.time()
                print("Epoch {}/{}".format(e, epochs),
                      "Iteration: {}".format(iteration),
                      "Avg. Training loss: {:.4f}".format(loss/100),
                      "{:.4f} sec/batch".format((end-start)/100))
                loss = 0
                start = time.time()
            
            # 计算相似的词
            if iteration % 1000 == 0:
                # 计算similarity
                sim = similarity.eval()
                for i in range(valid_size):
                    valid_word = int_to_vocab[valid_examples[i]]
                    top_k = 8 # 取最相似单词的前8个
                    nearest = (-sim[i, :]).argsort()[1:top_k+1]
                    log = 'Nearest to [%s]:' % valid_word
                    for k in range(top_k):
                        close_word = int_to_vocab[nearest[k]]
                        log = '%s %s,' % (log, close_word)
                    print(log)
            
            iteration += 1
            
    save_path = saver.save(sess, "checkpoints/text8.ckpt")
    embed_mat = sess.run(normalized_embedding)

8、可视化 

%matplotlib inline
%config InlineBackend.figure_format = 'retina'

import matplotlib.pyplot as plt
from sklearn.manifold import TSNE

viz_words = 500
tsne = TSNE()
embed_tsne = tsne.fit_transform(embed_mat[:viz_words, :])

fig, ax = plt.subplots(figsize=(14, 14))
for idx in range(viz_words):
    plt.scatter(*embed_tsne[idx, :], color='steelblue')
    plt.annotate(int_to_vocab[idx], (embed_tsne[idx, 0], embed_tsne[idx, 1]), alpha=0.7)

 

参考文献：http://localhost:8888/notebooks/Code_Lib/TensorFlow_jupyter/NLP/DeepLearning_21examples/chapter_14_word2vec/Word2vec_skipgram.ipynb