duongngyn0510 / contrastive-learning-prodlda-vinai

Contrast Learning experiment for Neural Topic Model in Classification task (5 classes)

Original paper

Result (Random Forest Classifiction with the number of decision trees as 30 and the maximum depth as 8 to predict the class of each document)

vector	f1-score
tfidf vector	0.671 +- 0.014
CLProdLDA latent vector	0.725 +- 0.005