Neural Network

Colab link

Colab link solved

This notebook demonstrates the process of performing text classification on text data, using the AG News dataset as an example.

Here’s a breakdown of the key concepts covered:

• Text Cleaning:
  • Converting text to lowercase.
  • Removing extra whitespaces, special characters, HTML or XML tags, and punctuation.
  • Expanding contractions.
  • Removing numbers and stop words.
  • Lemmatizing words.

• Data Splitting (training, validation and test): Dividing the dataset into training, validation, and test sets for model training and evaluation.

• Vectorization (Feature Extraction): Using CountVectorizer to convert text into numerical representations based on word counts. Using TF-IDF vectorizer to weigh word importance based on frequency.

• Model Building: Creating a simple neural network with a softmax activation function for classification. Compiling the model using an appropriate optimizer, loss function, and metrics.

• Model Training and Evaluation: Training the model on the training data and evaluating its performance on the test data using metrics like accuracy.