Towards Robust Named Entity Recognition for Historic German

Based on our paper we release a new model trained on the ONB dataset.

Note: We use BPEmbeddings instead of the combination of Wikipedia, Common Crawl and character embeddings (as used in the paper), so save space and training/inferencing time.

Results

Dataset \ Run Run 1 Run 2 Run 3 Avg.
Development 86.69 86.13 87.18 86.67
Test 85.27 86.05 85.75† 85.69

Paper reported an averaged F1-score of 85.31.

† denotes that this model is selected for upload.

Downloads last month
10
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.