Table 3 Macro-average F1 results on different datasets (%).
Model | MR | Ohsumed | R52 | R8 |
|---|---|---|---|---|
BERT | \(85.50\pm 0.21\) | \(62.50\pm 0.10\) | \(85.30\pm 0.15\) | \(95.05\pm 0.19\) |
XLNet | \(86.01\pm 0.20\) | \(62.67\pm 0.15\) | \(85.22\pm 0.14\) | \(95.10\pm 0.17\) |
SGC | \(71.88\pm 0.18\) | \(61.27\pm 0.23\) | \(74.18\pm 0.17\) | \(93.73\pm 0.21\) |
TextGCN | \(76.53\pm 0.18\) | \(62.15\pm 0.35\) | \(67.75\pm 0.21\) | \(93.83\pm 0.27\) |
TextING | \(79.66\pm 0.17\) | \(61.31\pm 0.32\) | \(73.32\pm 0.17\) | \(93.60\pm 0.22\) |
GLTC | \(80.20\pm 0.27\) | \(61.98\pm 0.24\) | \(74.81\pm 0.17\) | \(95.66\pm 0.18\) |
RB-GAT | \(79.13\pm 0.21\) | \(62.48\pm 0.32\) | \(74.11\pm 0.14\) | \(95.71\pm 0.15\) |
BertGCN | \(86.19\pm 0.22\) | \(62.17\pm 0.11\) | \(85.91\pm 0.26\) | \(95.76\pm 0.14\) |
XLG-Net | \({\textbf {88.55}}\pm {\textbf {0.22}}\) | \({\textbf {63.32}}\pm {\textbf {0.34}}\) | \({\textbf {86.45}}\pm {\textbf {0.31}}\) | \({\textbf {96.00}}\pm {\textbf {0.26}}\) |
