Skip to content

Development of a program to automatically detect malicious comments and classify them by category using Deep Learning

Notifications You must be signed in to change notification settings

B-logg/Malicious_Comment_Classification

Repository files navigation

악성 댓글 자동 감지 및 카테고리별 분류 프로그램 개발

사용 모델: klue/bert-base model
사용한 학습 데이터 셋: ‘Smilegate AI’에서 공개한 한국어 혐오 표현 “UnSmile” 데이터 셋(11 labels)
사용한 실험 데이터 셋: Korean HateSpeech Dataset

학습 결과

image
[6번째 모델 채택]

image
[test loss, Macro F1 Score, test set prediction result]

image
[정답과 예측 비교]

image
[정밀도, 재현율, f1-score]

실험 결과

image
[예측 결과 분포]

image

참고문헌

https://www.inews24.com/view/1444614
https://github.com/ZIZUN/korean-malicious-comments-dataset
https://github.com/kocohub/korean-hate-speech
https://github.com/smilegate-ai/korean_unsmile_dataset
https://wikidocs.net/31379
https://wikidocs.net/152922
https://wikidocs.net/166796
https://huggingface.co/klue/bert-base

About

Development of a program to automatically detect malicious comments and classify them by category using Deep Learning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors