UniNER-7B-all
| Property | Value |
|---|---|
| License | CC BY-NC 4.0 |
| Research Paper | arXiv:2308.03279 |
| Primary Language | English |
| Framework | PyTorch |
What is UniNER-7B-all?
UniNER-7B-all is a state-of-the-art named entity recognition model that represents the pinnacle of the UniNER model series. It's specifically designed for advanced NER tasks and has been trained on a comprehensive combination of ChatGPT-generated data and supervised datasets.
Implementation Details
The model combines three distinct data splits in its training: Pile-NER-type data, Pile-NER-definition data (both generated by ChatGPT), and carefully curated samples from 40 supervised datasets from the Universal NER benchmark. The implementation uses a sophisticated prompting template for inference, allowing for precise entity recognition.
- Built on LLaMA architecture with 7B parameters
- Trained on multiple high-quality data sources
- Implements text-generation-inference capabilities
- Supports entity-type specific queries
Core Capabilities
- Advanced named entity recognition across multiple domains
- JSON-formatted output for structured entity extraction
- Single entity type processing per query
- Research-oriented design with high accuracy
Frequently Asked Questions
Q: What makes this model unique?
The model's uniqueness lies in its comprehensive training approach, combining ChatGPT-generated data with supervised datasets, while excluding CrossNER and MIT datasets for OOD evaluation purposes.
Q: What are the recommended use cases?
This model is primarily designed for research applications in named entity recognition tasks, particularly when precise entity identification is required. It's especially useful for academic and research contexts where structured entity extraction is needed.
