Swedish-NER
| Property | Value |
|---|---|
| Developer | Recorded Future & AI Sweden |
| Base Model | KB/bert-base-swedish-cased |
| Model Type | Named Entity Recognition |
| Language | Swedish |
| HuggingFace URL | RecordedFuture/Swedish-NER |
What is Swedish-NER?
Swedish-NER is a specialized Named Entity Recognition model developed by Recorded Future in collaboration with AI Sweden. Built upon the KB/bert-base-swedish-cased architecture, this model is specifically designed for detecting and classifying named entities in Swedish text. The model has been fine-tuned using diverse data collected from various internet sources and forums, making it robust for real-world applications.
Implementation Details
The model is optimized for Transformers version 4.3.3 and above, running on Torch 1.8.0. It demonstrates impressive performance metrics across different entity categories, with an overall F1-score of 0.92. The model is specifically designed for Swedish language inputs and may not perform reliably on non-Swedish texts.
- Built on BERT-base architecture with Swedish language specialization
- Fine-tuned on diverse internet and forum data
- Supports six distinct entity categories
- Requires Transformers ≥4.3.3 and Torch 1.8.0
Core Capabilities
- Location detection (F1: 0.91)
- Organization identification (F1: 0.88)
- Person name recognition (F1: 0.96)
- Religion entity detection (F1: 0.91)
- Title recognition (F1: 0.84)
- Overall performance: F1-score of 0.92
Frequently Asked Questions
Q: What makes this model unique?
This model is specifically optimized for Swedish language NER tasks, with exceptional performance across multiple entity categories. Its training on diverse internet sources makes it particularly robust for real-world applications.
Q: What are the recommended use cases?
The model is ideal for Swedish text analysis tasks including document processing, information extraction, and content analysis where entity recognition is crucial. It's particularly useful for applications requiring identification of locations, organizations, persons, religions, and titles in Swedish text.
