Research on the Popularization of Marxism by Big Data Based on Attention Mechanism

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 7.

Figure 8.

Figure 8.

Comparison of accuracy indexes of ablation experiments(%)

Dimension	Bi-LSTM	CNN+ATT
valence	76.65	73.14
arousal	70.15	69.03

Comparison of specific indicators

Vector	Precision	Recall	F1
CNN+Attention(Random indicators)	0.8761	0.8483	0.8619
CNN+Attention(Specific indicators)	0.9176	0.8875	0.9021

Depth residual model structure with different layers

lay	out sizes	18-lay	34-lay	50-lay	101-lay	152-lay
convl	112x112	7x7,64,stride 2
		3x3 max pool, stride 2
conv2_x	56x56	[ 3×3,643x3,64 ]×2	[ 3×3,643×3,64 ]×3	[ 1x1,643x3,641x1,256 ]x3	[ 1×1,643×3,643×3,256 ]×3	[ 3×3,643×3,643×3,256 ]x3
conv3_x	28x28	[ 3×3,1283×3,128 ]×2	[ 3×3,1283×3,128 ]×4	[ 1×1,1283×3,1281×1,512 ]×4	[ 1×1,1283×3,1281×1,512 ]x4	[ 1x1,1283x3,1281x1,512 ]x8
conv4_x	14x14	[ 3x3,2563x3,256 ]x2	[ 3x3,2563x3,256 ]x6	[ 1x1,2563x3,2561x1,1024 ]x6	[ 1x1,2563x3,2561x1,1024 ]x23	[ 1×1,2563×3,2561x1,1024 ]x36
conv5_x	7x7	[ 3×3,5123×3,512 ]×2	[ 3x3,5123x3,512 ]x3	[ 1x1,5123x3,5121x1,2048 ]x3	[ 1×1,5123×3,5121×1,2048 ]x3	[ 1×1,5123x3,5121×1,2048 ]x3
	1x1	average pool,1000-d fc, softmax
FLOPs		1.8x109	3.6x109	3.8x109	7.6x109	11.3x109

Draw accuracy of neural network classification on sample 1 and sample 2 datasets

Dataset	Model	Accuracy	Standard deviation
Sample 1	DGCNN	90.04%	19.25
	DBN	86.08%	15.35
	CNN+ATT	93.73%	3.56
Sample 2	DGCNN	69.88%	25.68
	DBN	69.08%	35.21
	CNN+ATT	83.59%	8.92

Comparison of efficiency of different model experimental systems

Model	Precision	Recall	F1
SVM	0.9018	0.8743	0.8878
LR	0.7833	0.7642	0.7736
LSTM	0.9384	0.9353	0.9367
BiLSTM	0.9408	0.9321	0.9363
CNN	0.9264	0.9264	0.8641
CNN+ATT	0.9532	0.9401	0.9449