KLUE NER Baseline

Named Entity Recognition Baseline for KLUE Benchmark dataset

Baseline Metric

Training config available at config.yaml

Evaluation	Loss	Accuracy	Precision	Recall	Character F1 Score
klue/roberta-large with multisample dropout	0.069	0.981	0.901	0.915	0.908
klue/roberta-large	0.061	0.981	0.893	0.909	0.899
monologg/koelectra-base-v3-discriminator	0.085	0.976	0.853	0.891	0.872
kpfbert	0.086	0.972	0.846	0.877	0.861
klue/roberta-base	0.183	0.952	0.782	0.834	0.807

How to run

1. Change configurations in config.yaml
2. Run python3 main.py

Reference

• Huggingface Token Classification
• KLUE Dataset on Huggingface