Fake-EmoReact 2021 Challenge (Student)

https://competitions.codalab.org/competitions/31180?secret_key=2f97f399-8bba-4ed5-a0b7-99e17df1fe1b#participate

model

BERT-Base (L-12_H-768_A-12) + dropout + classifier

Confusion_Matrix

testing

method	accurracy	precision	recall
TF-IDF + Xgboost	0.966	0.971	0.82
GloVe + LSTM	0.971	0.976	0.989
GloVe + Bi-LSTM	0.971	0.971	0.989
BERT-Mini + dropout + classifier	0.971	0.976	0.989
BERT-Base + dropout + classifier	0.993	0.997	0.998

eval

BERT-Base (L-12_H-768_A-12)

• Precision score : 0.7394(18)
• Recall score : 0.6664(13)
• F1 score: 0.6353(14)

Top 14 in class.

Detial

https://github.com/linzino7/Fake-EmoReact-2021-Challenge/blob/main/NLP_Fakenews.pdf