How can Tensorflow be used to train the Illiad dataset using Python?

TensorFlow is a machine learning framework provided by Google. It is an open-source framework used with Python to implement algorithms, deep learning applications and much more. It is used for research and production purposes. It has optimization techniques that help perform complicated mathematical operations quickly by using NumPy and multi-dimensional arrays, also known as tensors.

The tensorflow package can be installed on Windows using the below command −

pip install tensorflow

We will be using the Iliad dataset, which contains text data of three translation works from William Cowper, Edward (Earl of Derby), and Samuel Butler. The model is trained to identify the translator when a single line of text is given. The text files have been preprocessed by removing document headers, footers, line numbers and chapter titles.

Dataset Configuration and Model Training

The following code demonstrates training a text classification model on the Iliad dataset. The model learns to identify which translator wrote a given line of text −

vocab_size += 2

print("Configure the dataset for better performance")
train_data = configure_dataset(train_data)
validation_data = configure_dataset(validation_data)

print("Train the model")
model = create_model(vocab_size=vocab_size, num_labels=3)
model.compile(
    optimizer='adam',
    loss=losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=['accuracy'])

print("Fit the training data to the model")
history = model.fit(train_data, validation_data=validation_data, epochs=3)

print("Finding the accuracy and loss associated with training")
loss, accuracy = model.evaluate(validation_data)

print("The loss is : ", loss)
print("The accuracy is : {:2.2%}".format(accuracy))

The output of the training process shows −

Configure the dataset for better performance
Train the model
Fit the training data to the model
Epoch 1/3
697/697 [==============================] - 35s 17ms/step - loss: 0.6891 - accuracy: 0.6736 -
val_loss: 0.3718 - val_accuracy: 0.8404
Epoch 2/3
697/697 [==============================] - 8s 11ms/step - loss: 0.3149 - accuracy: 0.8713 -
val_loss: 0.3621 - val_accuracy: 0.8422
Epoch 3/3
697/697 [==============================] - 8s 11ms/step - loss: 0.2165 - accuracy: 0.9162 -
val_loss: 0.4002 - val_accuracy: 0.8404
Finding the accuracy and loss associated with training
79/79 [==============================] - 1s 2ms/step - loss: 0.4002 - accuracy: 0.8404
The loss is : 0.40021833777427673
The accuracy is : 84.04%

How the Training Process Works

The training process involves several key steps −

• Dataset Configuration: The dataset is optimized for better performance using batching and prefetching techniques.

• Model Creation: A neural network model is created with the vocabulary size and number of labels (3 translators).

• Model Compilation: The model is compiled with Adam optimizer and sparse categorical crossentropy loss.

• Training: The model is trained for 3 epochs, showing improving accuracy from 67% to 91% on training data.

• Evaluation: Final validation accuracy reaches 84.04% with a loss of 0.40.

Key Training Metrics

Code credit − https://www.tensorflow.org/tutorials/load_data/text

Conclusion

The TensorFlow model successfully learns to classify text by translator with 84% validation accuracy. The training shows good convergence over 3 epochs, demonstrating TensorFlow's effectiveness for natural language processing tasks.