ner-english-ontonotes-large
| Property | Value |
|---|---|
| Research Paper | FLERT: Document-Level Features for Named Entity Recognition |
| Framework | Flair/PyTorch |
| Performance | 90.93% F1-Score on Ontonotes |
| Downloads | 159,630 |
What is ner-english-ontonotes-large?
The ner-english-ontonotes-large is a state-of-the-art Named Entity Recognition (NER) model designed for English text processing. Built on the FLERT architecture and utilizing document-level XLM-R embeddings, this model can identify and classify 18 different types of entities with high accuracy.
Implementation Details
The model is implemented using the Flair framework and leverages transformer-based architectures. It uses document-level contextual embeddings from XLM-RoBERTa-large and implements the FLERT approach for enhanced NER performance. The model architecture excludes CRF and RNN layers, focusing instead on direct fine-tuning of transformer embeddings.
- Utilizes XLM-RoBERTa-large as base model
- Implements document-level context for improved accuracy
- Fine-tunable embeddings with first subtoken pooling
- Trained using AdamW optimizer with OneCycleLR scheduler
Core Capabilities
- Recognizes 18 distinct entity types including PERSON, ORGANIZATION, DATE, etc.
- Processes document-level context for improved entity recognition
- Achieves 90.93% F1-score on the Ontonotes dataset
- Handles complex entity types like WORK_OF_ART and MONEY with high accuracy
Frequently Asked Questions
Q: What makes this model unique?
This model stands out due to its implementation of the FLERT architecture, which utilizes document-level features for enhanced NER performance. The combination of XLM-RoBERTa embeddings with document context allows for more accurate entity recognition across a wide range of entity types.
Q: What are the recommended use cases?
The model is ideal for applications requiring comprehensive entity recognition in English text, such as information extraction systems, content analysis tools, and document processing pipelines. It's particularly useful when dealing with diverse entity types beyond just names and locations.
