metadata
license: cc-by-nc-4.0
metrics:
- accuracy
- f1
datasets:
- universal_dependencies
language:
- eu
pipeline_tag: token-classification
tags:
- POS tagging
ElhBERTeu model finetuned for POS tagging on Universal Dependencies 1.2 corpus (eiec).
Results on test set:
- accuracy = 0.9793222286042504
- f1 = 0.9712429365605912
- loss = 0.12006692268620504
- precision = 0.9708653494328576
- recall = 0.9716208175035475
Tagset: [ADJ, ADP, ADV, AUX, CONJ, DET, INTJ, NOUN, NUM, PART, PRON, PROPN, PUNCT, SYM, VERB, X]
Finetuning details: 10 epochs, batch size 32, learning rate 3e-5.