Arabic-NER
| Property | Value |
|---|---|
| Parameter Count | 110M parameters |
| Framework Support | PyTorch, TensorFlow, JAX |
| Model Type | Token Classification (BERT-based) |
| Training Data | 378,000 tokens (14,000 sentences) |
What is arabic-ner?
Arabic-NER is a specialized Named Entity Recognition model built on the arabic-bert-base architecture, designed specifically for Arabic text analysis. This model has been trained to recognize nine distinct types of entities: PERSON, ORGANIZATION, LOCATION, DATE, PRODUCT, COMPETITION, PRIZE, EVENT, and DISEASE.
Implementation Details
The model is based on the arabic-bert-base architecture and has been trained on a manually annotated corpus of 378,000 tokens. It achieves an impressive F-measure of approximately 87% on validation data consisting of 30,000 tokens, demonstrating its robust performance in Arabic entity recognition tasks.
- Built on arabic-bert-base architecture
- Supports multiple deep learning frameworks including PyTorch, TensorFlow, and JAX
- Uses safetensors for model storage
- Implements token classification for precise entity detection
Core Capabilities
- Accurate identification of 9 distinct entity types
- Robust performance with 87% F-measure accuracy
- Handles complex Arabic text structures
- Support for both modern standard Arabic and dialectal variations
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its comprehensive coverage of Arabic named entities, supporting nine different entity types and achieving high accuracy. It's particularly noteworthy for its large training corpus of manually annotated Arabic text.
Q: What are the recommended use cases?
The model is ideal for Arabic text analysis tasks including information extraction, content categorization, news analysis, and automated document processing where entity recognition is crucial. It's particularly useful for applications requiring identification of persons, organizations, locations, and other entity types in Arabic text.
