Model

Dataset

Training Loss

Test Accuracy

Test Recall

Test F1 Score

Proposed

Military domain data

0.003

86.20%

83.10%

84.60%

BiDAF l

Military domain data

0.002

80.70%

77.30%

78.90%

R-Net

Military domain data

0.001

81.50%

78.60%

79.90%

XLNet

Military domain data

0.001

84.30%

81.90%

83.10%

Proposed

SQuAD dataset

0.002

80.50%

76.40%

78.20%

BiDAF

SQuAD dataset

0.004

76.20%

72.80%

74.30%

R-Net

SQuAD dataset

0.003

77.80%

74.20%

75.90%

XLNet

SQuAD dataset

0.005

82.40%

78.90%

80.60%