This project implements a Siamese neural network using the XLM-RoBERTa transformer to handle multilingual text data, significantly improving cross-lingual text classification tasks.
- Multilingual Text Handling: Utilizes XLM-RoBERTa transformer for processing and tokenizing large-scale multilingual datasets.
- Feature Engineering: Handles sequences up to 512 tokens, ensuring robust input representation.
- Transfer Learning: Adapts pre-trained language models for toxic comment classification across diverse linguistic contexts.
- High Performance: Achieved an accuracy of 92.5%, precision of 91.3%, recall of 90.8%, and an F1-score of 91.0% on the test dataset.
The model demonstrated the following performance on the test dataset:
- Accuracy: 92.5%
- Precision: 91.3%
- Recall: 90.8%
- F1-score: 91.0%
Contributions are welcome! Please open an issue or submit a pull request.
- Thanks to the developers of XLM-RoBERTa for their powerful multilingual transformer model.
- Special thanks to the open-source community for providing valuable resources and tools.