Text Classification Using CNN with TensorFlow

This project focuses on classifying text into categories using a Convolutional Neural Network (CNN) implemented in TensorFlow. The dataset used is a Persian text corpus, which undergoes preprocessing steps such as tokenization, stemming, lemmatization, and normalization. The model is trained using word embeddings and evaluated for its performance in classifying texts.

Requirements

Install the necessary libraries by running the following commands:

pip install hazm
pip install nltk
pip install tensorflow
pip install pandas
pip install matplotlib
pip install sklearn
pip install mlxtend

Data Preprocessing

1. Text Normalization: The Persian texts are normalized using the hazm library.
2. Stemming and Lemmatization: The texts are further processed by reducing words to their root forms.
3. Tokenization: The text is tokenized, converting each document into a sequence of words.
4. One-Hot Encoding: Labels are one-hot encoded for training.

Model Architecture

The model is a Sequential CNN with the following layers:

• Embedding Layer: Converts text sequences into dense vectors of fixed size.
• Conv1D Layer: Applies 1D convolutions to the embedded sequences.
• GlobalMaxPooling1D Layer: Reduces the dimensionality by taking the maximum value from each feature map.
• Dropout Layers: Adds regularization to prevent overfitting.
• Dense Output Layer: Outputs the classification in 34 categories using softmax activation.

Training

The model is trained using the following hyperparameters:

• Epochs: 5
• Batch size: 45
• Loss function: Categorical cross-entropy
• Optimizer: Adam

Callbacks

The model uses the following callbacks for optimization:

• Early Stopping
• Reduce Learning Rate on Plateau

Results Visualization

The model's accuracy and loss during training are plotted using matplotlib for both the training and validation datasets.

Prediction on Test Data

The model is used to predict the categories of a test dataset. The output is saved to a CSV file (out.csv).