HelsinkiNLP-FineTuned-Legal-es-zh
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Framework | PyTorch 1.8.1 |
| Task | Legal Text Translation (ES-ZH) |
| Training Data | 9,972 sentence pairs |
What is HelsinkiNLP-FineTuned-Legal-es-zh?
This is a specialized neural machine translation model fine-tuned for legal domain translation between Spanish and Chinese. Developed as part of a master's thesis at the Autonomous University of Barcelona, it builds upon the Helsinki-NLP's opus-tatoeba-es-zh base model with specific optimization for legal content.
Implementation Details
The model was trained using a carefully curated dataset of 9,972 sentence pairs from various legal sources, including the Spanish Civil Code, Spanish Constitution, and other regulations. Training utilized Adam optimizer with a learning rate of 2e-05, mixed precision training, and employed early stopping with a patience of 8 epochs.
- Training batch size: 8
- Validation split: 1,000 sentences
- Training epochs: 10
- Best validation loss: 2.0905 (step 5600)
Core Capabilities
- Specialized legal document translation between Spanish and Chinese
- Optimized for formal legal terminology and structures
- Supports both direct and inverse translation directions
- Particularly effective for legislative and regulatory content
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in legal domain translation, specifically trained on authentic legal documents from both Spanish and Chinese jurisdictions, making it particularly suited for professional legal translation tasks.
Q: What are the recommended use cases?
The model is primarily intended for academic and research purposes, as stated in the thesis objectives. It's particularly suitable for translating legal documents, regulations, and legislative texts between Spanish and Chinese.
