span-marker-mbert-base-multinerd
| Property | Value |
|---|---|
| Parameter Count | 178M |
| License | CC-BY-NC-SA 4.0 |
| F1 Score | 92.48% |
| Languages | Multilingual |
What is span-marker-mbert-base-multinerd?
This is a sophisticated multilingual Named Entity Recognition (NER) model that leverages the SpanMarker architecture with BERT-base-multilingual-cased as its foundation. The model excels at identifying 15 different types of entities across multiple languages, achieving an impressive overall F1 score of 92.48%. It's particularly noteworthy for its comprehensive entity coverage, ranging from person names and organizations to more specialized categories like celestial bodies and mythological entities.
Implementation Details
The model utilizes a span-based approach for entity recognition, trained on the MultiNERD dataset using the SpanMarker framework. It was trained for a single epoch with a learning rate of 5e-05 and batch size of 32, incorporating linear learning rate scheduling with 0.1 warmup ratio.
- Architecture: BERT-base-multilingual-cased encoder
- Training Framework: SpanMarker 1.2.4 with PyTorch 1.13.1
- Optimization: Adam optimizer with custom warmup scheduling
Core Capabilities
- Supports 15 entity types including PER, ORG, LOC, ANIM, BIO, CEL, DIS, EVE, FOOD, INST, MEDIA, PLANT, MYTH, TIME, VEHI
- Multilingual support with strong performance across 10 languages
- High precision (93.39%) and recall (91.59%) metrics
- Optimized for cased text processing
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its comprehensive multilingual capabilities and extensive entity type coverage. Unlike many NER models that focus on traditional entity types, it can recognize specialized categories like celestial bodies and mythological entities across multiple languages.
Q: What are the recommended use cases?
The model is ideal for multilingual text analysis, information extraction, and content classification tasks. It's particularly suitable for applications requiring detailed entity recognition across diverse domains like scientific literature, news articles, and cultural content analysis.
