Building a Discourse-Argument Hybrid System for Vietnamese Why-Question Answering

Table 1

The results of some deep learning models on SQuAD development set v1.1.

Model	F1
	All questions	Why-questions
R-NET+ (ensemble)	88.48%	66.90%
SLQA+ (ensemble)	88.38%	65.69%
Match-LSTM (boundary + ensemble)	76.76%	56.95%
BERT (ensemble)	92.2%	69.66%