bert-base-turkish-ner-cased
| Property | Value |
|---|---|
| Parameter Count | 111M parameters |
| Tensor Type | F32 |
| Research Paper | arXiv:2401.17396 |
| Downloads | 1,052 |
What is bert-base-turkish-ner-cased?
This is a specialized BERT-based model fine-tuned for Named Entity Recognition (NER) in Turkish text. Developed by Savas Yildirim, it achieves impressive F1 scores of 92.5% on standard Turkish NER tasks. The model is trained using transfer learning techniques on the WikiAnn dataset and implements state-of-the-art transformer architecture specifically optimized for Turkish language understanding.
Implementation Details
The model is implemented using the Transformers library and PyTorch backend. It utilizes a cased BERT architecture, meaning it maintains case sensitivity, which is crucial for NER tasks. Training was conducted with specific hyperparameters including a maximum sequence length of 128, batch size of 32, and 3 training epochs.
- Fine-tuned on Turkish WikiAnn dataset
- Implements full precision (F32) computation
- Supports inference endpoints for production deployment
- Achieves 91.6% precision and 93.4% recall on evaluation data
Core Capabilities
- Named Entity Recognition in Turkish text
- Support for complex Turkish language patterns
- Easy integration with Hugging Face Transformers pipeline
- Handles both modern and historical Turkish text formats
Frequently Asked Questions
Q: What makes this model unique?
This model is specifically optimized for Turkish NER tasks, maintaining case sensitivity and achieving state-of-the-art performance metrics. Its architecture is fine-tuned to handle Turkish language nuances effectively.
Q: What are the recommended use cases?
The model is ideal for applications requiring Turkish named entity recognition, such as information extraction, text analysis, and automated content categorization in Turkish documents. It's particularly effective for identifying person names, locations, and organizations in Turkish text.
