Toxicity Classifier

Methods Used

Preprocessing

• Text Vectorization of the entire data from TensorFlow
• MCSHBAP: map, cache, shuffle, batch, prefetch (to prepare dataset)

Building Model

• Sequential Model (one input, one output for each layer)
• Multiple Dense Layers to understand complex patterns (bidirectional lstm, relu, sigmoid)
• Compile the Model
• Save the Model

Prediction

• Load the Model
• Vectorize Input Text
• Predict using Model

Dataset Used

Kaggle Dataset Link

To DO

• Visualization (Corelation matrix, Result Toxicity percentage, acc - validation acc graph)
• Implementing Bert

Categorization of Work

• Visualization
• PreProcessing for LSTM
• LSTM Model
• Prediction using LSTM Model
• PreProcessing for Bert
• BERT Model
• Prediction using Bert Model

Work Division

• Krrish (Visualization, Predictions for LSTM and Bert)
• Nilay (PreProcessing for LSTM, LSTM Model)
• Shivansh (PreProcessing for Bert, Bert Model)