Comment Toxicity Analysis

Project Description

This project focuses on analyzing and detecting toxic comments using machine learning techniques, specifically a Long Short-Term Memory (LSTM) neural network. It aims to automatically identify and classify potentially harmful or offensive content in online discussions, helping to maintain healthier online communities.

Features

• Toxicity detection in comments using LSTM model
• Data preprocessing and tokenization
• Training, validation, and test dataset splitting
• Interactive testing interface using Gradio

Project Structure

COMMENT TOXICITY/
│
├── comment_toxicity_analysis/
│   └── __init__.py
│
├── data/
│   ├── sample_submission.csv
│   ├── test_labels.csv
│   ├── test.csv
│   └── train.csv
│
├── models/
│   ├── .gitkeep
│   └── toxicity.h5
│
├── notebooks/
│   ├── .gitkeep
│   └── comment_toxicity.ipynb
│
├── reports/
│   └── videos/
│
├── .env
├── .gitignore
├── pyproject.toml
├── README.md
├── requirements.txt
└── setup.cfg

Installation

1. Clone this repository:

   git clone https://github.com/chdl17/comment-toxicity.git
   cd comment-toxicity
   

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
   

3. Install the required packages:

   pip install -r requirements.txt
   

Usage

1. Data Preparation:

   • The project uses the train.csv file located in the data/ directory.
   • The script preprocesses the data, tokenizes the text, and prepares it for the model.

2. Model Training:

   • The model is a sequential LSTM neural network built using TensorFlow and Keras.
   • To train the model, run the provided Python script or Jupyter notebook.

3. Testing:

   • The project uses Gradio for interactive testing of the model.
   • Run the Gradio interface to input comments and see the toxicity analysis results.

Model Architecture

The core of this project is an LSTM (Long Short-Term Memory) model with the following architecture:

1. Embedding Layer
2. Bidirectional LSTM Layer
3. Dense Layers for feature extraction
4. Final Dense Layer with sigmoid activation for multi-label classification

The model is compiled using the Adam optimizer and binary crossentropy loss function.

Sample Code

Here's a snippet of the model architecture:

model = Sequential()
model.add(Embedding(max_words+1, 32))
model.add(Bidirectional(LSTM(32, activation='tanh')))
model.add(Dense(128, activation='relu'))
model.add(Dense(256, activation='relu'))
model.add(Dense(128, activation='relu'))
model.add(Dense(6, activation='sigmoid'))

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

Future Improvements

• Increase training data and epochs for better model performance
• Implement more advanced NLP techniques like attention mechanisms
• Develop a web application for easier access to the toxicity analysis tool
• Explore other model architectures for comparison

Technologies Used

• Python
• TensorFlow and Keras for deep learning
• Pandas for data manipulation
• Gradio for interactive testing
• Jupyter Notebook for development and analysis

Contact

Project Link: https://github.com/chdl17/comment-toxicity