📌 모델 상세 정보

1. 🧾 개요

• 클래스 구성:
  • "0": insult
  • "1": abuse
  • "2": obscenity
  • "3": TVPC(Threats of violence/promotion of crime)
  • "4": sexuality
  • "5": age
  • "6": race and region
  • "7": disabled
  • "8": religion
  • "9": politics
  • "10": job
  • "11": no_hate

2. 🧠 학습 정보

• Base Model: KrBERT-Medium (a pre-trained Korean language model based on BERT)
• Source: snunlp/KR-Medium
• Model Type: Casual Language Model
• Pre-training (Korean): 약 12.37GB (consisting of 91M and 1.17B words)
• Fine-tuning (Hate Dataset): 약 22.3MB (TTA-DQA/hate_sentence)
• Learning Rate: 5e-6
• Weight Decay: 0.01
• Epochs: 50
• Batch Size: 16
• Data Loader Workers: 2
• Tokenizer: BertWordPieceTokenizer
• Model Size: 약 405MB

3. 🧩 요구사항

• pytorch ~= 1.8.0
• transformers ~= 4.0.0

4. 🚀 Quick Start

5.📚 Citation

6. ⚠️ Bias, Risks, and Limitations

❗ 본 모델의 결과는 절대적인 유해 표현 기준이 아님을 유의해 주세요.

📈 Results

• Task: binary classification (text-classification)
• F1-score: 0.99005
• Accuracy: 0.99005