基于图自监督对比学习的社交媒体谣言检测

乔禹涵, 贾彩燕

Rumor detection on social media based on graph contrastive self⁃supervised learning

Yuhan Qiao, Caiyan Jia

表6 对比学习结合不同图编码器的实验结果

Table 6 Experimental results of contrastive learning by various graph encoders

模型 Twitter15 Twitter16 PHEME 准确率 Δ 准确率 Δ 准确率 Δ UDGAT 0.846 — 0.871 — 0.844 — UDGAT⁃GCSL 0.880 ↑3.4% 0.889 ↑1.8% 0.856 ↑1.2% BiGCN 0.843 — 0.858 — 0.847 — BiGCN⁃GCSL 0.881 ↑3.8% 0.888 ↑3.0% 0.850 ↑0.3% ClaHi⁃GAT 0.859 — 0.882 — 0.846 — ClaHi⁃GAT⁃GCSL 0.872 ↑1.3% 0.892 ↑1.0% 0.852 ↑0.6%