Tolerblanc/klue-bert-finetuned

K-urse_Detection_with_BERT

Overview

K-urse_Detection_with_BERT : Korean Cursing expression Detection with fine-tuned klue_BERT

This is the KWU "text mining" output for the first semester of 2023.

See Project Overview Here! : Notion(Korean)

See this model on GitHub : Link

Evaluation

• Comparison Model is here
• Used 2runo's Curse-detection-data

Model/Metric	Accuracy	Precision	Recall	F1 Score
Comparison(Electra base)	0.81	0.69	0.87	0.77
klue-BERT base(Our best result)	0.83	0.76**	0.75	0.75

• Used Youtube Comments

Model/Metric	Accuracy	Precision	Recall	F1 Score
Comparison(Electra base)	0.77	0.52	0.90	0.66
klue-BERT base(Our best result)	0.89	0.75	0.80	0.78

Demo with HuggingFace's Space 🤗

Try Demo Here! Go to HuggingFace Space

Reference

• Smilegate-AI's Korean Unsmile Dataset : Link
• JeanLee's K-MHaS Dataset and Paper : Link
• KLUE(Korean Language Understanding Evaluation) BERT : Link
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding : Link
• 2runo's Curse Detection Dataset : Link
• JminJ's Bad Text Classifier : Link