Copyright 2020 The TensorFlow Authors.

Multiclass text classification solution



In [ ]:

    
#@title Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

Overview

This notebook contains a solution to the multiclass classification exercise from the Basic text classification tutorial on tensorflow.org. This notebook contains contains few comments. To learn more, please refer to the tutorial.



In [ ]:

    
!pip install tf-nightly



In [ ]:

    
import tensorflow as tf

from tensorflow.keras import layers
from tensorflow.keras import losses
from tensorflow.keras import preprocessing
from tensorflow.keras.layers.experimental.preprocessing import TextVectorization



In [ ]:

    
print(tf.__version__)

Download the BigQuery dataset



In [ ]:

    
url = "http://storage.googleapis.com/download.tensorflow.org/data/stack_overflow_16k.tar.gz"

tf.keras.utils.get_file("stack_overflow_16k.tar.gz", url,
                        untar=True, cache_dir='.',
                        cache_subdir='')



In [ ]:

    
batch_size = 32

raw_train_ds = tf.keras.preprocessing.text_dataset_from_directory(
    'train', batch_size=batch_size, validation_split=0.2, subset='training', seed=42)

raw_val_ds = tf.keras.preprocessing.text_dataset_from_directory(
    'train', batch_size=batch_size, validation_split=0.2, subset='validation', seed=42)

raw_test_ds = tf.keras.preprocessing.text_dataset_from_directory(
    'test', batch_size=batch_size)

Explore the data



In [ ]:

    
print(raw_train_ds.class_names)



In [ ]:

    
for text_batch, label_batch in raw_train_ds.take(1):
  for i in range(5):
    print(text_batch.numpy()[i])
    print(label_batch.numpy()[i])
    print()

Prepare data for training



In [ ]:

    
max_features = 5000
embedding_dim = 128
sequence_length = 500

vectorize_layer = TextVectorization(
    max_tokens=max_features,
    output_mode='int',
    output_sequence_length=sequence_length)



In [ ]:

    
# Make a text-only dataset (no labels) and call adapt
text_ds = raw_train_ds.map(lambda x, y: x)
vectorize_layer.adapt(text_ds)

Vectorize the data



In [ ]:

    
def vectorize_text(text, label):
  text = tf.expand_dims(text, -1)
  return vectorize_layer(text), label

train_ds = raw_train_ds.map(vectorize_text)
val_ds = raw_val_ds.map(vectorize_text)
test_ds = raw_test_ds.map(vectorize_text)

Configure the dataset for performance



In [ ]:

    
AUTOTUNE = tf.data.experimental.AUTOTUNE

train_ds = train_ds.cache().prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)
test_ds = test_ds.cache().prefetch(buffer_size=AUTOTUNE)

Build the model

This model has not be tuned in any way, the goal is to show you the mechanics.



In [ ]:

    
model = tf.keras.Sequential([
  layers.Embedding(max_features + 1, embedding_dim),
  layers.Dropout(0.2),
  layers.GlobalAveragePooling1D(),
  layers.Dropout(0.2),
  layers.Dense(4)])



In [ ]:

    
model.compile(
    loss=losses.SparseCategoricalCrossentropy(from_logits=True), 
    optimizer='adam', 
    metrics=['accuracy'])

Train the model



In [ ]:

    
history = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=5)

Evaluate the model



In [ ]:

    
loss, accuracy = model.evaluate(test_ds)

print("Loss: ", loss)
print("Accuracy: ", accuracy)