text classification with keras in rt-polaritydata.

method	f1
navie bayes	0.7645+
FastText	0.7793+
TextCNN	0.7688+
TextRNN	0.7771+
RCNN	0.77
hierarchical attention network(HAN)	0.78+
Transformer("attention is all you need")	0.7636+
Bert	0.8135+