MODELS

ℹ️ Model Overview

Name	de_ent_depatent01 & de_ent_depatent02
Language	German (de)
Pipeline	ner
Authors	Bergeaud and Verluise
Date (last**)	02/2020
License	MIT

👷 Training

FORMAT=depatent01  # depatent02
spacy train configs/de_t2vner.cfg --paths.train data/train_ent_${FORMAT}.spacy --paths.dev data/train_ent_${FORMAT}.spacy --output models/de_ent_${FORMAT}

🔮 Model Performance

de_ent_depatent01/model-best

	ALL	ASG	CLAS	INV	LOC	OCC
p	0.99	0.98	0.99	0.99	1	0.97
r	0.99	0.99	1	0.96	1	0.98
f	0.99	0.98	1	0.98	1	0.97

de_ent_depatent02/model-best

	ALL	ASG	CIT	CLAS	INV	LOC	OCC
p	0.99	0.99	0	0.99	0.98	0.99	0.97
r	0.98	0.98	0	1	0.99	0.98	0.97
f	0.98	0.98	0	0.99	0.99	0.98	0.97

🎯 Intended use

de_ent_depatent0*_sm have been specifically trained on DE patents (resp DE1C-DE977922C and DE1000001B-). The model's performances are not guaranteed out of this scope.

🔂 Versions and alternative approaches

Version	Comment
0.1	ner - spaCy v2
1.0	ner - spaCy v3