作业五是根据Text8的语料库训练一个语言模型word2vec，得到语料库中每个词的嵌入式表达（向量）。
Mikolov提出的word2vec包括skip-gram和CBOW两种模型，前者是根据给定词预测其周围的词，后者是根据周围的词预测中间的词。Mikolov最大的贡献是采用negative sampling的方法极大提升神经网络模型的计算效率。官方的作业ipynb给出了skip-gram的实现，要求我们给出CBOW实现。
语料库
语料库使用Text8（http://mattmahoney.net/dc/tex8.zip），以下代码可以下载语料，但有可能因为网络原因没下完整，建议用浏览器下载完整。
import os
from six.moves.urllib.request import urlretrieve
url = 'http://mattmahoney.net/dc/'
def maybe_download(filename, expected_bytes):
"""Download a file if not present, and make sure it's the right size."""
if not os.path.exists(filename):
filename, _ = urlretrieve(url + filename, filename)
statinfo = os.stat(filename)
if statinfo.st_size == expected_bytes:
print('Found and verified %s' % 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('text8.zip', 31344016)注释：
下载完Text8.zip解压缩，读出数据并分词，大概1700w个重复词。设置字典大小vocabulary_size为5w，即只有5w个unique词，统计词频构建字典：
collections.Counter(words).most_common(vocabulary_size-1) 这个函数好强大，直接返回给定words列表的出现频数Top-K统计结果（词，词频），因为考虑要未知词UNKnown word（UNK），留了一个位置，K设置为字典大小减一。
根据统计词频count，构建字典dictionary，word作为key，word在count中的次序index作为value
遍历一遍词袋words，对词袋中的每个词，依次记录其在字典的索引index，得到data
最后的zip操作（构建tuple）是互换原来字典的key和value，重新构建一个字典reverse_dictionary，key是index，value是对应的词，方便根据索引找到词
vocabulary_size = 50000
def build_dataset(words):
count = [['UNK', -1]] # count的第一项
count.extend(collections.Counter(words).most_common(vocabulary_size - 1))
dictionary = dict()
for word, _ in count:
dictionary[word] = len(dictionary)
# key为word，value表示word在count中的次序index
data = list()
unk_count = 0
for word in words:
if word in dictionary:
index = dictionary[word]
else:
index = 0 # dictionary['UNK']
unk_count = unk_count + 1
data.append(index)
# data记录words每个词在字典中的index；不在字典的视为UNK，index用0
count[0][1] = unk_count # 用最新统计的unk词频取代原来('unk', -1)的-1
reverse_dictionary = dict(zip(dictionary.values(), dictionary.keys()))
# 互换原来字典的key和value，重新构建字典，方便根据index找到word
return data, count, dictionary, reverse_dictionary
data, count, dictionary, reverse_dictionary = build_dataset(words)
print('Most common words (+UNK)', count[:5])
print('Sample data', data[:10])
del words # Hint to reduce memory.skip-gram
可以先学习下skip-gram的基于TensorFlow的实现源码，关键在于构建训练数据&labels，以及使用tf.nn.embedding_lookup。为了区分原来注释，本人添加的注释为中文。
data_index = 0
def generate_batch(batch_size, num_skips, skip_window):
"""
batch_size, 每次训练数据的batch大小，是num_skips的整数倍，这里用8
num_skips, 目标词左和右的总词数，最大设为2倍skip_window
skip_window, 目标词左（右）的词数，计算整个窗口span=2*skip_window+1
"""
global data_index
assert batch_size % num_skips == 0
assert num_skips <= 2 * skip_window
# 初始化
batch = np.ndarray(shape=(batch_size), dtype=np.int32)
labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32)
span = 2 * skip_window + 1 # [ skip_window target skip_window ]
buffer = collections.deque(maxlen=span)
for _ in range(span): # 初始化一个span大小的buffer
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data) # 处理词序列边界，保证data_index在data范围
for i in range(batch_size // num_skips):
target = skip_window # target label at the center of the buffer
targets_to_avoid = [ skip_window ] # skip_window刚好是一个span中的中间词下标
for j in range(num_skips): # 遍历目标词的上下文，作为labels
while target in targets_to_avoid:
target = random.randint(0, span - 1) # 选上下文词：随机挑选span中除了target的一个词
targets_to_avoid.append(target) # 放入已选中列表，避免重复选
# batch中的词和labels一一对应，相同的target对应多个上下文词label
batch[i * num_skips + j] = buffer[skip_window]
labels[i * num_skips + j, 0] = buffer[target]
# 更新buffer，滑动窗口往后移动一个词，bufer是deque类型，追加一个词会弹出前面一个词
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
return batch, labels
print('data:', [reverse_dictionary[di] for di in data[:8]]) # 根据data索引di返回对应的词
# 测试不同的(num_skips, skip_window)，产生的batch和labels
for num_skips, skip_window in [(2, 1), (4, 2)]:
data_index = 0
batch, labels = generate_batch(batch_size=8, num_skips=num_skips, skip_window=skip_window)
print('\nwith num_skips = %d and skip_window = %d:' % (num_skips, skip_window))
print(' batch:', [reverse_dictionary[bi] for bi in batch])
print(' labels:', [reverse_dictionary[li] for li in labels.reshape(8)])注释：
skip-grams模型是给定中心词target预测其上下文词。在batch训练数据中，skip_window=1，则取target词的左右各1个词作为输出label，一个target对应多个label，因此看到batch出现2次”originated”，对应label分别为它的上下文”anarchism”和”as”，同理”as”, “a”, “term”亦如此。
skip_window=2时，取target词左右各2个词作为输出label，”as”对应前面2个词”anarchism”、”originated”和后面2个词”a”、”term”，因此batch出现4个相同的as，对应4个不同的label。
因此word2vec不用人工标注数据，依赖上下文数据，某次样本的输入同时也作为前后两次样本的输出，间接实现了监督学习。
batch_size = 128
embedding_size = 128 # Dimension of the embedding vector.
skip_window = 1 # How many words to consider left and right.
num_skips = 2 # 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.array(random.sample(range(valid_window), valid_size))
num_sampled = 64 # Number of negative examples to sample.
graph = tf.Graph()
with graph.as_default(), tf.device('/cpu:0'):
# Input data.
train_dataset = 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)
# Variables.
embeddings = tf.Variable(
tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0)) # embeddings：词库大小*向量维度
softmax_weights = tf.Variable(
tf.truncated_normal([vocabulary_size, embedding_size],
stddev=1.0 / math.sqrt(embedding_size))) # softmax的weights：词库大小*向量维度
softmax_biases = tf.Variable(tf.zeros([vocabulary_size])) # softmax的biases：词库大小
# Model.
# Look up embeddings for inputs.
embed = tf.nn.embedding_lookup(embeddings, train_dataset) # 按照train_dataset样本的ids顺序返回embeddings中的第ids行
# Compute the softmax loss, using a sample of the negative labels each time.
loss = tf.reduce_mean(
tf.nn.sampled_softmax_loss(weights=softmax_weights, biases=softmax_biases, inputs=embed,
labels=train_labels, num_sampled=num_sampled, num_classes=vocabulary_size))
# Optimizer.
# Note: The optimizer will optimize the softmax_weights AND the embeddings.
# This is because the embeddings are defined as a variable quantity and the
# optimizer's `minimize` method will by default modify all variable quantities
# that contribute to the tensor it is passed.
# See docs on `tf.train.Optimizer.minimize()` for more details.
optimizer = tf.train.AdagradOptimizer(1.0).minimize(loss)
# Compute the similarity between minibatch examples and all embeddings.
# We use the cosine distance:
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, tf.transpose(normalized_embeddings))CBOW
"""
For CBOW, given contextual words to predict central target word,
so use bag_window instead of num_skips & skip_window
"""
data_index = 0
def generate_batch(batch_size, bag_window):
"""
batch_size: 同skip-grams
bag_window: 输入待预测中心词的上（下）文词个数
"""
global data_index
# === modified ===
span = 2 * bag_window + 1 # [ bag_window target bag_window ]
batch = np.ndarray(shape=(batch_size, span - 1), dtype=np.int32)
labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32)
# ================
buffer = collections.deque(maxlen=span)
for _ in range(span):
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
for i in range(batch_size):
buffer_list = list(buffer)
labels[i, 0] = buffer_list[bag_window] # buffer的中间词下标是bag_window，作为要预测的label
batch[i] = np.append(buffer_list[:bag_window], buffer_list[bag_window+1:])
# 更新buffer，往下平移一个单词
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
return batch, labels
print('data_index:', data[:8])
print('data:', [reverse_dictionary[di] for di in data[:8]])
for bag_window in [1, 2]:
data_index = 0
batch, labels = generate_batch(batch_size=8, bag_window=bag_window)
print('\nwith bag_window = %d:' % (bag_window))
print(' batch:', [[reverse_dictionary[w] for w in bi] for bi in batch])
print(' labels:', [reverse_dictionary[li] for li in labels.reshape(8)])测试结果：
data_index: [5239, 3084, 12, 6, 195, 2, 3137, 46]
data: ['anarchism', 'originated', 'as', 'a', 'term', 'of', 'abuse', 'first']
with bag_window = 1:
batch: [['anarchism', 'as'], ['originated', 'a'], ['as', 'term'], ['a', 'of'], ['term', 'abuse'], ['of', 'first'], ['abuse', 'used'], ['first', 'against']]
labels: ['originated', 'as', 'a', 'term', 'of', 'abuse', 'first', 'used']
with bag_window = 2:
batch: [['anarchism', 'originated', 'a', 'term'], ['originated', 'as', 'term', 'of'], ['as', 'a', 'of', 'abuse'], ['a', 'term', 'abuse', 'first'], ['term', 'of', 'first', 'used'], ['of', 'abuse', 'used', 'against'], ['abuse', 'first', 'against', 'early'], ['first', 'used', 'early', 'working']]
labels: ['as', 'a', 'term', 'of', 'abuse', 'first', 'used', 'against']"""
CBOW Model
"""
batch_size = 128
embedding_size = 128 # Dimension of the embedding vector.
bag_window = 2 # 左右窗大小
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.array(random.sample(range(valid_window), valid_size))
num_sampled = 64 # Number of negative examples to sample.
graph = tf.Graph()
with graph.as_default(), tf.device('/cpu:0'):
# Input data.
train_dataset = tf.placeholder(tf.int32, shape=[batch_size, bag_window * 2]) # 在skip-grams基础上修改训练集输入
train_labels = tf.placeholder(tf.int32, shape=[batch_size, 1])
valid_dataset = tf.constant(valid_examples, dtype=tf.int32)
# Variables.
embeddings = tf.Variable(
tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0)) # embeddings：词库大小*向量维度
softmax_weights = tf.Variable(
tf.truncated_normal([vocabulary_size, embedding_size],
stddev=1.0 / math.sqrt(embedding_size))) # softmax的weights：词库大小*向量维度
softmax_biases = tf.Variable(tf.zeros([vocabulary_size])) # softmax的biases：词库大小
# Model.
# Look up embeddings for inputs.
embed = tf.nn.embedding_lookup(embeddings, train_dataset) # 按照train_dataset样本的ids顺序返回embeddings中的第ids行
# Compute the softmax loss, using a sample of the negative labels each time.
loss = tf.reduce_mean(
tf.nn.sampled_softmax_loss(weights=softmax_weights, biases=softmax_biases, inputs=tf.reduce_sum(embed, 1),
labels=train_labels, num_sampled=num_sampled, num_classes=vocabulary_size))
# 在skip-grams基础上修改inpus=tf.reduce_sum(embed, 1)
# Optimizer.
# Note: The optimizer will optimize the softmax_weights AND the embeddings.
# This is because the embeddings are defined as a variable quantity and the
# optimizer's `minimize` method will by default modify all variable quantities
# that contribute to the tensor it is passed.
# See docs on `tf.train.Optimizer.minimize()` for more details.
optimizer = tf.train.AdagradOptimizer(1.0).minimize(loss)
# Compute the similarity between minibatch examples and all embeddings.
# We use the cosine distance:
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, tf.transpose(normalized_embeddings))注释：
在skip-grams基础上，主要修改的地方有以下几个：
generate_batch函数，构造batch和label跟skip-grams相反，用bag_window代替num_skip和skip_window
tf.Graph()
train_dataset中，每个样本有2*bag_window维
损失函数的参数修改为inpus=tf.reduce_sum(embed, 1)，reduce_sum沿着1维度对tensor求和
num_steps = 100001
with tf.Session(graph=graph) as session:
tf.global_variables_initializer().run()
print('Initialized')
average_loss = 0
for step in range(num_steps):
batch_data, batch_labels = generate_batch(
batch_size, bag_window) # 修改参数为bag_windwow
feed_dict = {train_dataset : batch_data, train_labels : batch_labels}
_, l = session.run([optimizer, loss], feed_dict=feed_dict)
average_loss += l
if step % 2000 == 0:
if step > 0:
average_loss = average_loss / 2000
# The average loss is an estimate of the loss over the last 2000 batches.
print('Average loss at step %d: %f' % (step, average_loss))
average_loss = 0
# note that this is expensive (~20% slowdown if computed every 500 steps)
if step % 10000 == 0:
sim = similarity.eval()
for i in range(valid_size):
valid_word = reverse_dictionary[valid_examples[i]]
top_k = 8 # number of nearest neighbors
nearest = (-sim[i, :]).argsort()[1:top_k+1]
log = 'Nearest to %s:' % valid_word
for k in range(top_k):
close_word = reverse_dictionary[nearest[k]]
log = '%s %s,' % (log, close_word)
print(log)
final_embeddings = normalized_embeddings.eval()转自：http://blog.csdn.net/draco_mystack/article/details/77456604
源码参考：https://github.com/Zerof007/uda_deeplearning_z