Finnish NER Extended 2020
TurkuNLP publishes an extended Finnish NER archive named combined-ext-model-130220. It uses FinBERT and the 18-category OntoNotes scheme.
Research release
An extended entity scheme
The extended release covers more than the six entity types in the Turku NER corpus. It includes categories such as geopolitical entity, facility and work of art, alongside numerical and time expressions. The official project page links an archive that remains downloadable over HTTP; a working HTTPS archive URL was not verified.TurkuNLP Finnish NER: extended trained model and download
Separate code and weight terms
The linked Keras implementation is MIT-licensed. That software license does not by itself state the terms for the separately hosted trained model archive.Finnish BERT NER implementation
Keep labels distinct
Practical guidance. Map each output category deliberately to the destination schema. Merging locations, geopolitical entities and facilities can remove information that this extended model was trained to identify. Older Keras packaging may also require a separate conversion before use in a current pipeline.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | combined-ext-model-130220 |
| Label scheme | 18 OntoNotes categories, including names, places, works, laws, dates and quantities.TurkuNLP Finnish NER: extended trained model and download |
| Download | combined-ext-model-130220.tar.gz, linked by the official project page.TurkuNLP Finnish NER: extended trained model and download |
| Weight license | No explicit license for the trained archive was located on the reviewed project page.TurkuNLP Finnish NER: extended trained model and download |
Related models
- FinBERT Base Cased
TurkuNLP ยท FinBERT
FinBERT Base Cased is a Finnish masked-language encoder with a dedicated WordPiece vocabulary.
Launch date Downloadable