Data Source: Kaggle
Description: This end-to-end Deep Learning project trained a model to receive a comment and give a response of if it falls into any of the predefined categories.
The categories are:
- "toxic"
- "severe_toxic"
- "obscene"
- "threat"
- "insult"
- "identity_hate"
- "safe"
Approach: This was done using Bert as a base model.
Result: The model attained an accuracy of 92.97% after 5 epochs of training.
Deployment: Finally the model was deployed as an API endpoint using FastAPI.