Tensorflow-word2vec_simple.py的理解

Tensorflow-word2vec_simple.py的理解

1. 基本思想
   
   • 推荐word2vec图文理解
   • Word2vec Tutorial
2. 代码注释与理解
   我的代码及注释如下：

# -*- coding: utf-8 -*-
"""
Basic word2vec example
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import collections
import os
import random
import zipfile
import tensorflow as tf

import numpy as np
from six.moves import urllib
import urllib
# step1: download the data
url = 'http://mattmahoney.net/dc/'


def maybe_download(filename, expected_bytes):
    """
    Download a file if not present, and make sure it is the right size
    :param filename:
    :param expected_bytes:
    :return:
    """
    if not os.path.exists(filename):
        filename, _ = urllib.request.urlretrieve(url + filename, filename)
    statinfo = os.stat(filename)
    if statinfo.st_size == expected_bytes:
        print("found and verified", filename)
    else:
        print(statinfo.st_size)
        raise Exception('Failed to verify' + filename + 'Can you get to it with a browser?')
    return filename

filename = maybe_download('../../data/corpus/text8.zip', 31344016)

# read the data into a list of strings.
def read_data(filename):
    """
    Extract the first file enclosed in a zip file as a list of words.
    :param filename:
    :return:
    """
    with zipfile.ZipFile(filename) as f:
        data = tf.compat.as_str(f.read(f.namelist()[0])).split()
    return data

vocabulary = read_data(filename)
print('Data size', len(vocabulary))

# step 2: Build the dictionary and replace rare words with UNK token.
vocabulary_size = 50000

def build_dataset(words, n_words):
    """
    process raw input to dataset.
    构建输入数据以及形成了代码单词对照表，data将被用于训练模型，而dictionary将可以作为最后的查询矢量及单词关系的翻译本
    :param words: a list of words
    :param n_words: the number of words
    :return:
       'count': a list ,[['UNK', count], (word1, count1), (word2, count2), (word3, count3)...], most common n_words
       'dictionary': {word1: index1, word2: index2, word3: index3...}, most common n_words
       'data': a list, [index1, index2, index3, ...], all indexes of words and the index of 'UNK' is zero ,index表示词频排名
       'reversed_dictionary': a dictionary,  the reversed dictionary of 'dictionary'
    """
    # 词频统计
    count = [['UNK', -1]]
    count.extend(collections.Counter(words).most_common(n_words - 1))
    # 对count的词频进行整理，word作为key，排行顺序作为value
    dictionary = dict()
    for word, _ in count:
        dictionary[word] = len(dictionary)
    # 构建输入数据data,每一项为对应word的排名
    data = list()
    unk_count = 0
    for word in words:
        if word in dictionary:
            index = dictionary[word]
        else:
            index = 0  # dictionary['UNK]
            unk_count += 1
        data.append(index)
    count[0][1] = unk_count
    reversed_dictionary = dict(zip(dictionary.values(), dictionary.keys()))
    return data, count, dictionary, reversed_dictionary

data, count, dictionary, reversed_dictionary = build_dataset(vocabulary, vocabulary_size)

del vocabulary  # Hint to reduce memory
print('Most common words (+UNK)', count[: 5])
print('Sample data', data[: 10], [reversed_dictionary[i] for i in data[: 10]])

data_index = 0

# step3: Function to generate a training batch for the skip-gram model.
def generate_batch(batch_size, num_skips, skip_window):
    """
    生成word2vec训练样本
    :param batch_size: 每个批次训练样本的多少
    :param num_skips: How many times to reuse an input to generate a label
    :param skip_window: How many words to consider left and right
    :return:
      ‘batch‘:
     'labels':
    """
    global data_index
    assert batch_size % num_skips == 0
    assert num_skips <= 2 * skip_window
    batch = np.ndarray(shape=batch_size, dtype=np.int32)
    labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32)
    span = 2 * skip_window + 1  # 队列长度
    buffer = collections.deque(maxlen=span)  # 使用双端队列 buffer来存取w的上下文的id
    # 初始化
    for _ in range(span):
        buffer.append(data[data_index])
        data_index = (data_index + 1) % len(data)
    # 移动窗口获取批量数据
    for i in range(batch_size // num_skips):  # how many num_skips in a batch
        target = skip_window       # target label at the center of the buffer
        targets_to_avoid = [skip_window]  # extract the middle word
        for j in range(num_skips):
            while target in targets_to_avoid:  # context中的word，一个只取一次
                target = random.randint(0, span - 1)
            targets_to_avoid.append(target)
            batch[i * num_skips + j] = buffer[skip_window]
            labels[i * num_skips + j, 0] = buffer[target]
        buffer.append(data[data_index])  # update the buffer, append the next word to buffer
        data_index = (data_index + 1) % len(data)
    # Backtrack a little bit to avoid skipping words in the end of batch
    data_index = (data_index + len(data) - span) % len(data)
    return batch, labels

batch, labels = generate_batch(batch_size=8, num_skips=2, skip_window=1)
for i in range(8):
    print(batch[i], reversed_dictionary[batch[i]], '->', labels[i, 0], reversed_dictionary[labels[i, 0]])

# Step 4: Build and train a skip-gram model
batch_size = 128
embedding_size = 128  # Dimensions of the Embedding vector
skip_window = 3  # How many words to consider left and right
num_skips = 4  # How many times to reuse an input to generate a label

# We pick a random validation set to sample nearest neighbors. Here we limit the
# validation samples to the words that have a low numeric ID, which by
# construction are also the most frequent.
valid_size = 16     # Random set of words to evaluate similarity on.
valid_window = 100  # Only pick dev samples in the head of the distribution.
valid_examples = np.random.choice(valid_window, valid_size, replace=False)
num_sampled = 64    # Number of negative examples to sample.

graph = tf.Graph()

with graph.as_default():
    # input data
    train_inputs = tf.placeholder(tf.int32, shape=[batch_size])
    train_labels = tf.placeholder(tf.int32, shape=[batch_size, 1])
    valid_dataset = tf.constant(valid_examples, dtype=tf.int32)

    with tf.device('/cpu: 0'):
        # Look up embeddings for inputs
        embeddings = tf.Variable(tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0))
        embed = tf.nn.embedding_lookup(embeddings, train_inputs)

        # Construct the Variables for the NCE loss
        nce_weights = tf.Variable(tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0))
        nce_biases = tf.Variable(tf.zeros([vocabulary_size]))

    # compute the average NCE loss for the batch
    # tf.nce_loss automatically draws a new sample of the negative labels each time we evaluate the loss
    loss = tf.reduce_mean(tf.nn.nce_loss(weights=nce_weights,
                                         biases=nce_biases,
                                         labels=train_labels,
                                         inputs=embed,
                                         num_sampled=num_sampled,
                                         num_classes=vocabulary_size))
    optimizer = tf.train.GradientDescentOptimizer(0.8).minimize(loss)
    # compute the cosine similarity between minibatch examples  and all embeddings
    norm = tf.sqrt(tf.reduce_sum(tf.square(embeddings), 1, keep_dims=True))
    normalized_embeddings = embeddings / norm

    valid_embeddings = tf.nn.embedding_lookup(normalized_embeddings, valid_dataset)
    similarity = tf.matmul(valid_embeddings, normalized_embeddings, transpose_b=True)

    # add variable initializer
    init = tf.global_variables_initializer()

# step five: Begin training
num_steps = 100001

with tf.Session(graph=graph) as session:
    init.run()
    print("initialized")
    average_loss = 0
    for step in range(num_steps):
        batch_inputs, batch_labels = generate_batch(batch_size, num_skips, skip_window)
        feed_dict = {train_inputs: batch_inputs, train_labels: batch_labels}

        # perform one update step by evaluating the optimizer op
        _, loss_val = session.run([optimizer, loss], feed_dict=feed_dict)
        average_loss += loss_val

        if step % 2000 == 0:
            if step > 0:
                average_loss /= 2000
            # The average loss is an estimate of the loss over the last 2000 batches.
            print('Average loss at step ', step, ': ', average_loss)
            average_loss = 0

            # Note that this is expensive (~20% slowdown if computed every 500 steps)
            # 计算验证单词和全部单词的相似度，验证单词最相似8个单词展示
        if step % 10000 == 0:
            sim = similarity.eval()   # tf.Tensor.eval(feed_dict=None, session=None)
            for i in range(valid_size):
                valid_word = reversed_dictionary[valid_examples[i]]
                top_k = 8  # number of nearest neighbors
                nearest = (-sim[i, :]).argsort()[1:top_k + 1]   # 负号表示降序排列
                log_str = 'Nearest to %s:' % valid_word
                for k in range(top_k):
                    close_word = reversed_dictionary[nearest[k]]
                    log_str = '%s %s,' % (log_str, close_word)
                print(log_str)
    # execute this tensor in a session
    final_embeddings = normalized_embeddings.eval()

# step six: Visualize the embeddings
def plot_with_labels(low_dim_embs, labels, filename='tsne.png'):
    assert low_dim_embs.shape[0] >= len(labels), 'More labels than embeddings'
    plt.figure(figsize=(18, 18))  # in inches
    for i, label in enumerate(labels):
        x, y = low_dim_embs[i, :]
        plt.scatter(x, y)
        plt.annotate(label,
                     xy=(x, y),
                     xytext=(5, 2),
                     textcoords='offset points',
                     ha='right',
                     va='bottom')

    plt.savefig(filename)

# pylint: disable=g-import-not-at-top
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt

tsne = TSNE(perplexity=30, n_components=2, init='pca', n_iter=5000)
plot_only = 500
low_dim_embs = tsne.fit_transform(final_embeddings[:plot_only, :])  # 降到2维
labels = [reversed_dictionary[i] for i in range(plot_only)]
plot_with_labels(low_dim_embs, labels)

3. 推荐参考文章

• Distributed Representations of Words and Phrases
  and their Compositionality（Miklov.et 2013）