Sentiment Analysis on Movie Revies

In [0]:

    

!pip install -q tf-nightly-gpu-2.0-preview


    

In [2]:

    

import tensorflow as tf
print(tf.__version__)


    

    




2.0.0-dev20190502

In [0]:

    

import matplotlib.pyplot as plt
import pandas as pd
import tensorflow as tf


    

In [0]:

    

# https://keras.io/datasets/#imdb-movie-reviews-sentiment-classification
max_features = 10000  # number of words to consider as features
maxlen = 500  # cut texts after this number of words (among top max_features most common words)

# each review is encoded as a sequence of word indexes
# indexed by overall frequency in the dataset
# output is 0 (negative) or 1 (positive) 
imdb = tf.keras.datasets.imdb.load_data(num_words=max_features)
(raw_input_train, y_train), (raw_input_test, y_test) = imdb

# https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/sequence/pad_sequences

input_train = tf.keras.preprocessing.sequence.pad_sequences(raw_input_train, maxlen=maxlen)
input_test = tf.keras.preprocessing.sequence.pad_sequences(raw_input_test, maxlen=maxlen)


    

In [5]:

    

from tensorflow.keras.layers import GRU, Embedding, BatchNormalization, Dropout
from tensorflow.keras.models import Sequential, Model

embedding_dim = 32

dropout = 0.7

model = Sequential()

# encoder
model.add(Embedding(input_dim=max_features, output_dim=embedding_dim, input_length=maxlen))

# return_sequences passes all outputs of all timesteps (not only the last one) to the next layer
model.add(GRU(name='gru1', units=32, return_sequences=True))

# for embedding: 32*2 (“standard deviation” parameter (gamma), “mean” parameter (beta)) trainable parameters
# and 32*2 (moving_mean and moving_variance) non-trainable parameters
model.add(BatchNormalization())  
model.add(Dropout(dropout))

# stack recurrent layers like with fc
model.add(GRU(name='gru2', units=32))
model.add(BatchNormalization())
model.add(Dropout(dropout))

# latent space
model.add(tf.keras.layers.Dense(name='fc', units=32, activation='relu'))

# binary classifier as decoder
model.add(tf.keras.layers.Dense(name='classifier', units=1, activation='sigmoid'))

model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

model.summary()


    

    




WARNING: Logging before flag parsing goes to stderr.
W0502 19:06:56.682862 140566851819392 nn_ops.py:4202] Large dropout rate: 0.7 (>0.5). In TensorFlow 2.x, dropout() uses dropout rate instead of keep_prob. Please ensure that this is intended.
W0502 19:06:57.382216 140566851819392 nn_ops.py:4202] Large dropout rate: 0.7 (>0.5). In TensorFlow 2.x, dropout() uses dropout rate instead of keep_prob. Please ensure that this is intended.






    




Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
embedding (Embedding)        (None, 500, 32)           320000    
_________________________________________________________________
gru1 (GRU)                   (None, 500, 32)           6336      
_________________________________________________________________
batch_normalization (BatchNo (None, 500, 32)           128       
_________________________________________________________________
dropout (Dropout)            (None, 500, 32)           0         
_________________________________________________________________
gru2 (GRU)                   (None, 32)                6336      
_________________________________________________________________
batch_normalization_1 (Batch (None, 32)                128       
_________________________________________________________________
dropout_1 (Dropout)          (None, 32)                0         
_________________________________________________________________
fc (Dense)                   (None, 32)                1056      
_________________________________________________________________
classifier (Dense)           (None, 1)                 33        
=================================================================
Total params: 334,017
Trainable params: 333,889
Non-trainable params: 128
_________________________________________________________________

In [6]:

    

batch_size = 1000

%time history = model.fit(input_train, y_train, epochs=15, batch_size=batch_size, validation_split=0.4)


    

    




Train on 15000 samples, validate on 10000 samples
Epoch 1/15
15000/15000 [==============================] - 128s 9ms/sample - loss: 0.9416 - accuracy: 0.5109 - val_loss: 0.6932 - val_accuracy: 0.5033
Epoch 2/15
15000/15000 [==============================] - 128s 9ms/sample - loss: 0.7826 - accuracy: 0.5363 - val_loss: 0.6939 - val_accuracy: 0.5033
Epoch 3/15
15000/15000 [==============================] - 125s 8ms/sample - loss: 0.7103 - accuracy: 0.5753 - val_loss: 0.6936 - val_accuracy: 0.5033
Epoch 4/15
15000/15000 [==============================] - 126s 8ms/sample - loss: 0.6470 - accuracy: 0.6333 - val_loss: 0.6900 - val_accuracy: 0.5033
Epoch 5/15
15000/15000 [==============================] - 127s 8ms/sample - loss: 0.5690 - accuracy: 0.7126 - val_loss: 0.6802 - val_accuracy: 0.5043
Epoch 6/15
15000/15000 [==============================] - 126s 8ms/sample - loss: 0.4706 - accuracy: 0.7894 - val_loss: 0.6349 - val_accuracy: 0.5998
Epoch 7/15
15000/15000 [==============================] - 128s 9ms/sample - loss: 0.3616 - accuracy: 0.8592 - val_loss: 0.6244 - val_accuracy: 0.5795
Epoch 8/15
15000/15000 [==============================] - 125s 8ms/sample - loss: 0.2638 - accuracy: 0.9113 - val_loss: 0.6234 - val_accuracy: 0.5834
Epoch 9/15
15000/15000 [==============================] - 126s 8ms/sample - loss: 0.2033 - accuracy: 0.9351 - val_loss: 0.5449 - val_accuracy: 0.6892
Epoch 10/15
15000/15000 [==============================] - 126s 8ms/sample - loss: 0.1588 - accuracy: 0.9541 - val_loss: 0.4279 - val_accuracy: 0.8317
Epoch 11/15
15000/15000 [==============================] - 125s 8ms/sample - loss: 0.1298 - accuracy: 0.9638 - val_loss: 0.4055 - val_accuracy: 0.8367
Epoch 12/15
15000/15000 [==============================] - 126s 8ms/sample - loss: 0.1189 - accuracy: 0.9661 - val_loss: 0.4855 - val_accuracy: 0.7584
Epoch 13/15
15000/15000 [==============================] - 124s 8ms/sample - loss: 0.0826 - accuracy: 0.9811 - val_loss: 0.3884 - val_accuracy: 0.8323
Epoch 14/15
15000/15000 [==============================] - 124s 8ms/sample - loss: 0.0658 - accuracy: 0.9870 - val_loss: 0.4308 - val_accuracy: 0.8081
Epoch 15/15
15000/15000 [==============================] - 125s 8ms/sample - loss: 0.0621 - accuracy: 0.9858 - val_loss: 0.3984 - val_accuracy: 0.8308
CPU times: user 55min 14s, sys: 3min 28s, total: 58min 43s
Wall time: 31min 29s

In [7]:

    

import matplotlib.pyplot as plt

plt.ylabel("accuracy")
plt.xlabel("epochs")

plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])

plt.legend(["Accuracy", "Valdation Accuracy"])


    

    
Out[7]:





<matplotlib.legend.Legend at 0x7fd7d9375c88>






    

In [8]:

    

train_loss, train_accuracy = model.evaluate(input_train, y_train, batch_size=batch_size)
train_accuracy


    

    




25000/25000 [==============================] - 16s 659us/sample - loss: 0.2308 - accuracy: 0.9284






    
Out[8]:





0.92836

In [9]:

    

test_loss, test_accuracy = model.evaluate(input_test, y_test, batch_size=batch_size)
test_accuracy


    

    




25000/25000 [==============================] - 18s 703us/sample - loss: 0.4162 - accuracy: 0.8174






    
Out[9]:





0.81736

In [20]:

    

# precition
model.predict(input_test[0:5])


    

    
Out[20]:





array([[0.26075655],
       [0.97113156],
       [0.8640883 ],
       [0.24265748],
       [0.9815538 ]], dtype=float32)

In [11]:

    

# ground truth
y_test[0:5]


    

    
Out[11]:





array([0, 1, 1, 0, 1])