README.md

Dacon_Sentiment_Type_classification

Result

	F1	RANK
Public	0.7552	15
Private	0.75359	11

The main problem and how to solve it

• class 불균형

• loss 계산 비중 조정
• generalization
• Oversampling

• multi label text classification vs text classification for #4

• 2가지 모두 사용

..to be continue

Reference

Reference about loss function

• R-Drop: Regularized Dropout for Neural Networks (NeurlPS 2021)

  • Reference Github link

• SMART: Robust and Efficient Fine-Tuning for Pre-trained Natural Language Models through Principled Regularized Optimization (ACL 2020)

• Dual Contrastive Learning: Text Classification via Label-Aware Data Augmentation (2022)

• Balancing Methods for Multi-label Text Classification with Long-Tailed Class Distribution (EMNLP 2021)

Reference about Model

• An Algorithm for Routing Vectors in Sequences (2022)
  • Heinsen routing Algorithm