Hybrid Low-Order and Higher-Order Graph Convolutional Networks

Table 3

Text network classification accuracy.

Methods	R52	OH	20NG	R8	MR
CNN-rand [22]	87.59	58.44	82.15	95.71	77.75
LSTM [23]	85.54	41.13	65.71	93.68	75.06
LSTM-pre [23]	90.48	51.10	75.43	96.09	77.33
PTE [24]	90.71	53.58	76.74	96.69	70.23
fastText [25]	92.81	57.70	79.38	96.13	75.14
SWEM [26]	92.94	63.12	85.16	95.32	76.65
LEAM [27]	91.84	58.58	81.91	93.31	76.95
GCN-C [13]	92.75	63.86	81.42	96.99	77.22
GCN-S [5]	92.74	62.82	—	96.80	76.99
GCN-F [11]	93.20	63.04	—	96.89	76.74
Text GCN [21]	93.56	68.36	86.34	97.07	76.74
HLHG-2 (ours)	94.21 ± 0.14	69.16 ± 0.19	86.57 ± 0.08	97.25 ± 0.10	75.95 ± 0.14
HLHG-3 (ours)	94.33 ± 0.16	69.36 ± 0.24	86.35 ± 0.24	97.25 ± 0.12	76.49 ± 0.32