Trang này được dịch bởi Cloud Translation API.
Switch to English

Text classification with an RNN

View on TensorFlow.org Run in Google Colab View source on GitHub Download notebook

This text classification tutorial trains a recurrent neural network on the IMDB large movie review dataset for sentiment analysis.

Setup

pip install -q tfds-nightly
import tensorflow_datasets as tfds
import tensorflow as tf

Import matplotlib and create a helper function to plot graphs:

import matplotlib.pyplot as plt

def plot_graphs(history, metric):
  plt.plot(history.history[metric])
  plt.plot(history.history['val_'+metric], '')
  plt.xlabel("Epochs")
  plt.ylabel(metric)
  plt.legend([metric, 'val_'+metric])
  plt.show()

Setup input pipeline

The IMDB large movie review dataset is a binary classification dataset—all the reviews have either a positive or negative sentiment.

Download the dataset using TFDS.

dataset, info = tfds.load('imdb_reviews/subwords8k', with_info=True,
                          as_supervised=True)
train_dataset, test_dataset = dataset['train'], dataset['test']
WARNING:absl:TFDS datasets with text encoding are deprecated and will be removed in a future version. Instead, you should use the plain text version and tokenize the text using `tensorflow_text` (See: https://www.tensorflow.org/tutorials/tensorflow_text/intro#tfdata_example)

Downloading and preparing dataset imdb_reviews/subwords8k/1.0.0 (download: 80.23 MiB, generated: Unknown size, total: 80.23 MiB) to /home/kbuilder/tensorflow_datasets/imdb_reviews/subwords8k/1.0.0...
Shuffling and writing examples to /home/kbuilder/tensorflow_datasets/imdb_reviews/subwords8k/1.0.0.incompleteUNSS8I/imdb_reviews-train.tfrecord
Shuffling and writing examples to /home/kbuilder/tensorflow_datasets/imdb_reviews/subwords8k/1.0.0.incompleteUNSS8I/imdb_reviews-test.tfrecord

Warning:absl:Dataset is using deprecated text encoder API which will be removed soon. Please use the plain_text version of the dataset and migrate to `tensorflow_text`.

Shuffling and writing examples to /home/kbuilder/tensorflow_datasets/imdb_reviews/subwords8k/1.0.0.incompleteUNSS8I/imdb_reviews-unsupervised.tfrecord
Dataset imdb_reviews downloaded and prepared to /home/kbuilder/tensorflow_datasets/imdb_reviews/subwords8k/1.0.0. Subsequent calls will reuse this data.

The dataset info includes the encoder (a tfds.deprecated.text.SubwordTextEncoder).

encoder = info.features['text'].encoder
print('Vocabulary size: {}'.format(encoder.vocab_size))
Vocabulary size: 8185

This text encoder will reversibly encode any string, falling back to byte-encoding if necessary.

sample_string = 'Hello TensorFlow.'

encoded_string = encoder.encode(sample_string)
print('Encoded string is {}'.format(encoded_string))

original_string = encoder.decode(encoded_string)
print('The original string: "{}"'.format(original_string))
Encoded string is [4025, 222, 6307, 2327, 4043, 2120, 7975]
The original string: "Hello TensorFlow."

assert original_string == sample_string
for index in encoded_string:
  print('{} ----> {}'.format(index, encoder.decode([index])))
4025 ----> Hell
222 ----> o 
6307 ----> Ten
2327 ----> sor
4043 ----> Fl
2120 ----> ow
7975 ----> .

Prepare the data for training

Next create batches of these encoded strings. Use the padded_batch method to zero-pad the sequences to the length of the longest string in the batch:

BUFFER_SIZE = 10000
BATCH_SIZE = 64
train_dataset = train_dataset.shuffle(BUFFER_SIZE)
train_dataset = train_dataset.padded_batch(BATCH_SIZE)

test_dataset = test_dataset.padded_batch(BATCH_SIZE)

Create the model

Build a tf.keras.Sequential model and start with an embedding layer. An embedding layer stores one vector per word. When called, it converts the sequences of word indices to sequences of vectors. These vectors are trainable. After training (on enough data), words with similar meanings often have similar vectors.

This index-lookup is much more efficient than the equivalent operation of passing a one-hot encoded vector through a tf.keras.layers.Dense layer.

A recurrent neural network (RNN) processes sequence input by iterating through the elements. RNNs pass the outputs from one timestep to their input—and then to the next.

The tf.keras.layers.Bidirectional wrapper can also be used with an RNN layer. This propagates the input forward and backwards through the RNN layer and then concatenates the output. This helps the RNN to learn long range dependencies.

model = tf.keras.Sequential([
    tf.keras.layers.Embedding(encoder.vocab_size, 64),
    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(64)),
    tf.keras.layers.Dense(64, activation='relu'),
    tf.keras.layers.Dense(1)
])

Please note that Keras sequential model is used here since all the layers in the model only have single input and produce single output. In case you want to use stateful RNN layer, you might want to build your model with Keras functional API or model subclassing so that you can retrieve and reuse the RNN layer states. Please check Keras RNN guide for more details.

Compile the Keras model to configure the training process:

model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
              optimizer=tf.keras.optimizers.Adam(1e-4),
              metrics=['accuracy'])

Train the model

history = model.fit(train_dataset, epochs=10,
                    validation_data=test_dataset, 
                    validation_steps=30)
Epoch 1/10
391/391 [==============================] - 40s 103ms/step - loss: 0.6505 - accuracy: 0.5492 - val_loss: 0.4595 - val_accuracy: 0.7620
Epoch 2/10
391/391 [==============================] - 40s 104ms/step - loss: 0.3331 - accuracy: 0.8583 - val_loss: 0.3579 - val_accuracy: 0.8719
Epoch 3/10
391/391 [==============================] - 41s 104ms/step - loss: 0.2460 - accuracy: 0.9042 - val_loss: 0.3280 - val_accuracy: 0.8594
Epoch 4/10
391/391 [==============================] - 41s 104ms/step - loss: 0.2060 - accuracy: 0.9240 - val_loss: 0.3350 - val_accuracy: 0.8531
Epoch 5/10
391/391 [==============================] - 40s 103ms/step - loss: 0.1785 - accuracy: 0.9363 - val_loss: 0.3538 - val_accuracy: 0.8693
Epoch 6/10
391/391 [==============================] - 41s 104ms/step - loss: 0.1635 - accuracy: 0.9402 - val_loss: 0.3705 - val_accuracy: 0.8687
Epoch 7/10
391/391 [==============================] - 41s 104ms/step - loss: 0.1405 - accuracy: 0.9532 - val_loss: 0.4286 - val_accuracy: 0.8656
Epoch 8/10
391/391 [==============================] - 41s 104ms/step - loss: 0.1269 - accuracy: 0.9574 - val_loss: 0.3992 - val_accuracy: 0.8609
Epoch 9/10
391/391 [==============================] - 40s 104ms/step - loss: 0.1156 - accuracy: 0.9626 - val_loss: 0.4647 - val_accuracy: 0.8536
Epoch 10/10
391/391 [==============================] - 41s 105ms/step - loss: 0.1136 - accuracy: 0.9626 - val_loss: 0.4158 - val_accuracy: 0.8167

test_loss, test_acc = model.evaluate(test_dataset)

print('Test Loss: {}'.format(test_loss))
print('Test Accuracy: {}'.format(test_acc))
391/391 [==============================] - 16s 41ms/step - loss: 0.4312 - accuracy: 0.8018
Test Loss: 0.431190550327301
Test Accuracy: 0.801800012588501

The above model does not mask the padding applied to the sequences. This can lead to skew if trained on padded sequences and test on un-padded sequences. Ideally you would use masking to avoid this, but as you can see below it only have a small effect on the output.

If the prediction is >= 0.5, it is positive else it is negative.

def pad_to_size(vec, size):
  zeros = [0] * (size - len(vec))
  vec.extend(zeros)
  return vec
def sample_predict(sample_pred_text, pad):
  encoded_sample_pred_text = encoder.encode(sample_pred_text)

  if pad:
    encoded_sample_pred_text = pad_to_size(encoded_sample_pred_text, 64)
  encoded_sample_pred_text = tf.cast(encoded_sample_pred_text, tf.float32)
  predictions = model.predict(tf.expand_dims(encoded_sample_pred_text, 0))

  return (predictions)
# predict on a sample text without padding.

sample_pred_text = ('The movie was cool. The animation and the graphics '
                    'were out of this world. I would recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=False)
print(predictions)
[[0.01823657]]

# predict on a sample text with padding

sample_pred_text = ('The movie was cool. The animation and the graphics '
                    'were out of this world. I would recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=True)
print(predictions)
[[-0.10063186]]

plot_graphs(history, 'accuracy')

png

plot_graphs(history, 'loss')

png

Stack two or more LSTM layers

Keras recurrent layers have two available modes that are controlled by the return_sequences constructor argument:

  • Return either the full sequences of successive outputs for each timestep (a 3D tensor of shape (batch_size, timesteps, output_features)).
  • Return only the last output for each input sequence (a 2D tensor of shape (batch_size, output_features)).
model = tf.keras.Sequential([
    tf.keras.layers.Embedding(encoder.vocab_size, 64),
    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(64,  return_sequences=True)),
    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(32)),
    tf.keras.layers.Dense(64, activation='relu'),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(1)
])
model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
              optimizer=tf.keras.optimizers.Adam(1e-4),
              metrics=['accuracy'])
history = model.fit(train_dataset, epochs=10,
                    validation_data=test_dataset,
                    validation_steps=30)
Epoch 1/10
391/391 [==============================] - 74s 190ms/step - loss: 0.6606 - accuracy: 0.5480 - val_loss: 0.5329 - val_accuracy: 0.7203
Epoch 2/10
391/391 [==============================] - 75s 192ms/step - loss: 0.3841 - accuracy: 0.8429 - val_loss: 0.3507 - val_accuracy: 0.8599
Epoch 3/10
391/391 [==============================] - 75s 191ms/step - loss: 0.2715 - accuracy: 0.8994 - val_loss: 0.3651 - val_accuracy: 0.8714
Epoch 4/10
391/391 [==============================] - 75s 192ms/step - loss: 0.2295 - accuracy: 0.9172 - val_loss: 0.3576 - val_accuracy: 0.8635
Epoch 5/10
391/391 [==============================] - 75s 192ms/step - loss: 0.1906 - accuracy: 0.9353 - val_loss: 0.3610 - val_accuracy: 0.8651
Epoch 6/10
391/391 [==============================] - 75s 192ms/step - loss: 0.1652 - accuracy: 0.9475 - val_loss: 0.4046 - val_accuracy: 0.8677
Epoch 7/10
391/391 [==============================] - 75s 192ms/step - loss: 0.1447 - accuracy: 0.9573 - val_loss: 0.4394 - val_accuracy: 0.8568
Epoch 8/10
391/391 [==============================] - 76s 193ms/step - loss: 0.1243 - accuracy: 0.9657 - val_loss: 0.5010 - val_accuracy: 0.8406
Epoch 9/10
391/391 [==============================] - 75s 191ms/step - loss: 0.1127 - accuracy: 0.9692 - val_loss: 0.4892 - val_accuracy: 0.8531
Epoch 10/10
391/391 [==============================] - 75s 192ms/step - loss: 0.0948 - accuracy: 0.9764 - val_loss: 0.5652 - val_accuracy: 0.8505

test_loss, test_acc = model.evaluate(test_dataset)

print('Test Loss: {}'.format(test_loss))
print('Test Accuracy: {}'.format(test_acc))
391/391 [==============================] - 30s 76ms/step - loss: 0.5370 - accuracy: 0.8510
Test Loss: 0.5369970798492432
Test Accuracy: 0.8510400056838989

# predict on a sample text without padding.

sample_pred_text = ('The movie was not good. The animation and the graphics '
                    'were terrible. I would not recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=False)
print(predictions)
[[-2.3630884]]

# predict on a sample text with padding

sample_pred_text = ('The movie was not good. The animation and the graphics '
                    'were terrible. I would not recommend this movie.')
predictions = sample_predict(sample_pred_text, pad=True)
print(predictions)
[[-2.690089]]

plot_graphs(history, 'accuracy')

png

plot_graphs(history, 'loss')

png

Check out other existing recurrent layers such as GRU layers.

If you're interestied in building custom RNNs, see the Keras RNN Guide.