In this notebook, I'll lead you through using TensorFlow to implement the word2vec algorithm using the skip-gram architecture. By implementing this, you'll learn about embedding words for use in natural language processing. This will come in handy when dealing with things like translations.
Here are the resources I used to build this notebook. I suggest reading these either beforehand or while you're working on this material.
When you're dealing with language and words, you end up with tens of thousands of classes to predict, one for each word. Trying to one-hot encode these words is massively inefficient, you'll have one element set to 1 and the other 50,000 set to 0. The word2vec algorithm finds much more efficient representations by finding vectors that represent the words. These vectors also contain semantic information about the words. Words that show up in similar contexts, such as "black", "white", and "red" will have vectors near each other. There are two architectures for implementing word2vec, CBOW (Continuous Bag-Of-Words) and Skip-gram.
In this implementation, we'll be using the skip-gram architecture because it performs better than CBOW. Here, we pass in a word and try to predict the words surrounding it in the text. In this way, we can train the network to learn representations for words that show up in similar contexts.
First up, importing packages.
In [1]:
import time
import numpy as np
import tensorflow as tf
import utils
Load the text8 dataset, a file of cleaned up Wikipedia articles from Matt Mahoney. The next cell will download the data set to the data folder. Then you can extract it and delete the archive file to save storage space.
In [2]:
from urllib.request import urlretrieve
from os.path import isfile, isdir
from tqdm import tqdm
import zipfile
dataset_folder_path = 'data'
dataset_filename = 'text8.zip'
dataset_name = 'Text8 Dataset'
class DLProgress(tqdm):
last_block = 0
def hook(self, block_num=1, block_size=1, total_size=None):
self.total = total_size
self.update((block_num - self.last_block) * block_size)
self.last_block = block_num
if not isfile(dataset_filename):
with DLProgress(unit='B', unit_scale=True, miniters=1, desc=dataset_name) as pbar:
urlretrieve(
'http://mattmahoney.net/dc/text8.zip',
dataset_filename,
pbar.hook)
if not isdir(dataset_folder_path):
with zipfile.ZipFile(dataset_filename) as zip_ref:
zip_ref.extractall(dataset_folder_path)
with open('data/text8') as f:
text = f.read()
Here I'm fixing up the text to make training easier. This comes from the utils module I wrote. The preprocess function coverts any punctuation into tokens, so a period is changed to <PERIOD>. In this data set, there aren't any periods, but it will help in other NLP problems. I'm also removing all words that show up five or fewer times in the dataset. This will greatly reduce issues due to noise in the data and improve the quality of the vector representations. If you want to write your own functions for this stuff, go for it.
In [3]:
words = utils.preprocess(text)
print(words[30:35])
In [4]:
print("Total words: {}".format(len(words)))
print("Unique words: {}".format(len(set(words))))
And here I'm creating dictionaries to covert words to integers and backwards, integers to words. The integers are assigned in descending frequency order, so the most frequent word ("the") is given the integer 0 and the next most frequent is 1 and so on. The words are converted to integers and stored in the list int_words.
In [5]:
vocab_to_int, int_to_vocab = utils.create_lookup_tables(words)
In [6]:
print(vocab_to_int['the'])
print(int_to_vocab[3528])
In [7]:
len(int_to_vocab)
Out[7]:
In [8]:
int_words = [vocab_to_int[word] for word in words]
In [9]:
print(words[0:30])
print(int_words[0:30])
Words that show up often such as "the", "of", and "for" don't provide much context to the nearby words. If we discard some of them, we can remove some of the noise from our data and in return get faster training and better representations. This process is called subsampling by Mikolov. For each word $w_i$ in the training set, we'll discard it with probability given by
$$ P(w_i) = 1 - \sqrt{\frac{t}{f(w_i)}} $$where $t$ is a threshold parameter and $f(w_i)$ is the frequency of word $w_i$ in the total dataset.
I'm going to leave this up to you as an exercise. This is more of a programming challenge, than about deep learning specifically. But, being able to prepare your data for your network is an important skill to have. Check out my solution to see how I did it.
Exercise: Implement subsampling for the words in
int_words. That is, go throughint_wordsand discard each word given the probablility $P(w_i)$ shown above. Note that $P(w_i)$ is the probability that a word is discarded. Assign the subsampled data totrain_words.
In [10]:
int_words.count(3528)
Out[10]:
In [11]:
import random as rnd
np.random.rand()
Out[11]:
In [12]:
import random
from collections import Counter
## Your code here
train_words = []
words_counter = Counter(int_words)
total_number_words = len(int_words)
k = 0.00001
#freq = {word:count/tot_count for word,count in words_counter.items()}
prob_words = [1-np.sqrt(k/(words_counter[i]/total_number_words)) for i in range(len(vocab_to_int))]
for int_word in int_words:
if prob_words[int_word] < random.random():
train_words.append(int_word)
In [13]:
print(len(int_words))
print(len(train_words))
Now that our data is in good shape, we need to get it into the proper form to pass it into our network. With the skip-gram architecture, for each word in the text, we want to grab all the words in a window around that word, with size $C$.
From Mikolov et al.:
"Since the more distant words are usually less related to the current word than those close to it, we give less weight to the distant words by sampling less from those words in our training examples... If we choose $C = 5$, for each training word we will select randomly a number $R$ in range $< 1; C >$, and then use $R$ words from history and $R$ words from the future of the current word as correct labels."
Exercise: Implement a function
get_targetthat receives a list of words, an index, and a window size, then returns a list of words in the window around the index. Make sure to use the algorithm described above, where you choose a random number of words from the window.
In [14]:
def get_target(words, idx, window_size=1):
''' Get a list of words in a window around an index. '''
list_words =[]
# Your code here
R = random.randint(1,window_size+1)
start = idx - R if (idx - R) > 0 else 0
end = idx + R
set_words = set(words[start:idx] + words[idx+1:end+1] )
#range_test = [start:idx] + [idx+1:end+1]
#print(range_test, " with index = ", idx, " , R = " , R)
return list(set_words)
In [15]:
def get_target_edited(words, idx, window_size=5):
''' Get a list of words in a window around an index. '''
list_words =[]
# Your code here
R = random.randint(1,window_size)
R_2 = R
start = idx - R_2 if (idx - R_2) > 0 else 0
end = idx + R_2
idx_start = (idx - R if (idx - R) > 0 else 0)
idx_end = idx + R
range_test = list(range(idx+1,idx_end+1))
print(range_test, " with index = ", idx, " , R = " , R)
print("words correct: ", words[start:idx] + words[idx+1:end+1])
print("words incorrect: ", [words[i] for i in list(range(idx_start,idx))+list(range(idx+1,idx_end+1))])
#print(idx_end+1, " with R : " ,R)
return list(set([words[i] for i in list(range(idx_start,idx))+list(range(idx+1,idx_end+1))]))
In [16]:
get_target_edited(words,100)
Out[16]:
In [17]:
get_target(words,100)
Out[17]:
In [18]:
words[98:102]
Out[18]:
Here's a function that returns batches for our network. The idea is that it grabs batch_size words from a words list. Then for each of those words, it gets the target words in the window. I haven't found a way to pass in a random number of target words and get it to work with the architecture, so I make one row per input-target pair. This is a generator function by the way, helps save memory.
In [19]:
def get_batches(words, batch_size, window_size=5):
''' Create a generator of word batches as a tuple (inputs, targets) '''
n_batches = len(words)//batch_size
# only 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 ii in range(len(batch)):
batch_x = batch[ii]
batch_y = get_target(batch, ii, window_size)
y.extend(batch_y)
x.extend([batch_x]*len(batch_y))
yield x, y
From Chris McCormick's blog, we can see the general structure of our network.
The input words are passed in as one-hot encoded vectors. This will go into a hidden layer of linear units, then into a softmax layer. We'll use the softmax layer to make a prediction like normal.
The idea here is to train the hidden layer weight matrix to find efficient representations for our words. This weight matrix is usually called the embedding matrix or embedding look-up table. We can discard the softmax layer becuase we don't really care about making predictions with this network. We just want the embedding matrix so we can use it in other networks we build from the dataset.
I'm going to have you build the graph in stages now. First off, creating the inputs and labels placeholders like normal.
Exercise: Assign
inputsandlabelsusingtf.placeholder. We're going to be passing in integers, so set the data types totf.int32. The batches we're passing in will have varying sizes, so set the batch sizes to [None]. To make things work later, you'll need to set the second dimension oflabelstoNoneor1.
In [20]:
train_graph = tf.Graph()
with train_graph.as_default():
inputs = tf.placeholder(tf.int32, [None], name='inputs')
labels = tf.placeholder(tf.int32, [None, None], name='labels')
The embedding matrix has a size of the number of words by the number of neurons in the hidden layer. So, if you have 10,000 words and 300 hidden units, the matrix will have size $10,000 \times 300$. Remember that we're using one-hot encoded vectors for our inputs. When you do the matrix multiplication of the one-hot vector with the embedding matrix, you end up selecting only one row out of the entire matrix:
You don't actually need to do the matrix multiplication, you just need to select the row in the embedding matrix that corresponds to the input word. Then, the embedding matrix becomes a lookup table, you're looking up a vector the size of the hidden layer that represents the input word.
Exercise: Tensorflow provides a convenient function
tf.nn.embedding_lookupthat does this lookup for us. You pass in the embedding matrix and a tensor of integers, then it returns rows in the matrix corresponding to those integers. Below, set the number of embedding features you'll use (200 is a good start), create the embedding matrix variable, and usetf.nn.embedding_lookupto get the embedding tensors. For the embedding matrix, I suggest you initialize it with a uniform random numbers between -1 and 1 using tf.random_uniform. This TensorFlow tutorial will help if you get stuck.
In [21]:
n_vocab = len(int_to_vocab)
n_embedding = 200# Number of embedding features
with train_graph.as_default():
#tf.random_uniform((n_vocab, n_embedding), -1, 1)
embedding = tf.Variable(tf.random_uniform([n_vocab,n_embedding],-1,1), name = 'embedding') # create embedding weight matrix here
embed = tf.nn.embedding_lookup(embedding,inputs) # use tf.nn.embedding_lookup to get the hidden layer output
For every example we give the network, we train it using the output from the softmax layer. That means for each input, we're making very small changes to millions of weights even though we only have one true example. This makes training the network very inefficient. We can approximate the loss from the softmax layer by only updating a small subset of all the weights at once. We'll update the weights for the correct label, but only a small number of incorrect labels. This is called "negative sampling". Tensorflow has a convenient function to do this, tf.nn.sampled_softmax_loss.
Exercise: Below, create weights and biases for the softmax layer. Then, use
tf.nn.sampled_softmax_lossto calculate the loss. Be sure to read the documentation to figure out how it works.
In [22]:
# Number of negative labels to sample
n_sampled = 100
with train_graph.as_default():
softmax_w = tf.Variable(tf.random_normal([n_vocab,n_embedding],stddev=0.1))
softmax_b = tf.Variable(tf.zeros(n_vocab,dtype=tf.float32))
# Calculate the loss using negative sampling
loss = tf.nn.sampled_softmax_loss(softmax_w, softmax_b,
labels,embed,n_sampled,
n_vocab )
cost = tf.reduce_mean(loss)
optimizer = tf.train.AdamOptimizer().minimize(cost)
In [23]:
with train_graph.as_default():
## From Thushan Ganegedara's implementation
valid_size = 16 # Random set of words to evaluate similarity on.
valid_window = 100
# pick 8 samples from (0,100) and (1000,1100) each ranges. lower id implies more frequent
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_dataset = tf.constant(valid_examples, dtype=tf.int32)
# We use the cosine distance:
norm = tf.sqrt(tf.reduce_sum(tf.square(embedding), 1, keep_dims=True))
normalized_embedding = embedding / norm
valid_embedding = tf.nn.embedding_lookup(normalized_embedding, valid_dataset)
similarity = tf.matmul(valid_embedding, tf.transpose(normalized_embedding))
In [ ]:
# If the checkpoints directory doesn't exist:
!mkdir checkpoints
In [ ]:
epochs = 10
batch_size = 1000
window_size = 10
with train_graph.as_default():
saver = tf.train.Saver()
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()
if iteration % 1000 == 0:
## From Thushan Ganegedara's implementation
# note that this is expensive (~20% slowdown if computed every 500 steps)
sim = similarity.eval()
for i in range(valid_size):
valid_word = int_to_vocab[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 = 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)
Restore the trained network if you need to:
In [ ]:
with train_graph.as_default():
saver = tf.train.Saver()
with tf.Session(graph=train_graph) as sess:
saver.restore(sess, tf.train.latest_checkpoint('checkpoints'))
embed_mat = sess.run(embedding)
Below we'll use T-SNE to visualize how our high-dimensional word vectors cluster together. T-SNE is used to project these vectors into two dimensions while preserving local stucture. Check out this post from Christopher Olah to learn more about T-SNE and other ways to visualize high-dimensional data.
In [ ]:
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
import matplotlib.pyplot as plt
from sklearn.manifold import TSNE
In [ ]:
viz_words = 500
tsne = TSNE()
embed_tsne = tsne.fit_transform(embed_mat[:viz_words, :])
In [ ]:
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)