bert-large-NER
| Property | Value |
|---|---|
| Parameter Count | 334M parameters |
| License | MIT |
| Paper | Original BERT Paper |
| Performance | F1: 91.7%, Precision: 91.2%, Recall: 92.3% |
What is bert-large-NER?
bert-large-NER is a state-of-the-art Named Entity Recognition model based on the BERT-large architecture. Fine-tuned on the CoNLL-2003 dataset, it specializes in identifying four types of entities: Location (LOC), Organization (ORG), Person (PER), and Miscellaneous (MISC). The model represents a significant advancement in NER tasks, utilizing the power of transformer architecture with 334M parameters.
Implementation Details
The model was trained on a single NVIDIA V100 GPU using the recommended hyperparameters from the original BERT paper. It processes text using a cased vocabulary and maintains BERT's powerful contextual understanding while specializing in entity recognition.
- Built on bert-large-cased architecture
- Fine-tuned on CoNLL-2003 Named Entity Recognition dataset
- Supports both beginning (B-) and inside (I-) tags for entity recognition
- Achieves 91.7% F1 score on test set
Core Capabilities
- Accurate identification of person names, organizations, locations, and miscellaneous entities
- Support for identifying consecutive entities of the same type
- Maintains contextual understanding across long sequences
- Easy integration with Hugging Face's transformers pipeline
Frequently Asked Questions
Q: What makes this model unique?
The model combines BERT-large's powerful language understanding capabilities with specific optimization for NER tasks, achieving near state-of-the-art performance without requiring document context or CRF layers.
Q: What are the recommended use cases?
The model is ideal for applications requiring named entity extraction from news articles, documents, and general text analysis. It's particularly effective for identifying proper nouns and organizational names in formal text.
