donghoon-io/KoSentimentAnalysis

Training code for Korean multi-class sentiment analysis (Published in CHI 2022)

KoSentimentAnalysis

Bert implementation for the Korean multi-class sentiment analysis

왜 한국어 감정 다중분류 모델은 거의 없는 것일까?에서 시작된 프로젝트

ARM64 환경에 최적화

0. Environment: Pytorch, Dataset: 감성 대화 말뭉치
1. Install dependencies:

!pip install sentencepiece
!pip install torch

2. Populate data (we in this code assume 6-class classification tasks, based on Ekman's sentiment model)
3. Train
4. Validate & Use (See below # test comment)