tner-xlm-roberta-base-ontonotes5
| Property | Value |
|---|---|
| Author | Asahi Ushio |
| Model Type | Token Classification |
| Base Architecture | XLM-RoBERTa |
| Paper | Research Paper |
| Downloads | 18,395 |
| Language | English |
What is tner-xlm-roberta-base-ontonotes5?
This is a specialized Named Entity Recognition (NER) model built on the XLM-RoBERTa architecture and fine-tuned on the OntoNotes5 dataset. It features 12 attention heads and 12 hidden layers, with a vocabulary size of 250,002 tokens, making it particularly effective for multilingual token classification tasks.
Implementation Details
The model utilizes the transformer architecture with key specifications including a layer normalization epsilon of 1e-05 and is implemented using PyTorch. It's designed to work seamlessly with the tner library and can be easily integrated into existing NLP pipelines.
- 12 attention heads for enhanced context understanding
- 12 hidden layers for deep feature extraction
- Comprehensive token classification capabilities
- Support for multiple entity types including ORG, MISC, PER, and LOC
Core Capabilities
- Named Entity Recognition across multiple languages
- Token-level classification with 9 distinct label categories
- Integration with popular NLP frameworks
- Support for both direct and downstream applications
Frequently Asked Questions
Q: What makes this model unique?
This model combines the multilingual capabilities of XLM-RoBERTa with specialized NER training on OntoNotes5, making it particularly effective for cross-lingual named entity recognition tasks.
Q: What are the recommended use cases?
The model is ideal for tasks requiring named entity recognition in multilingual contexts, including organization name detection, person name identification, location recognition, and miscellaneous entity classification.
