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5 不同训练数据规模下的实验结果

Table 5 Experimental results with various scales of labeled training data

数据集 模型 10% 20% 50% 80% 准确率 Δ 准确率 Δ 准确率 Δ 准确率 Δ Twitter15 UDGAT 0.608 — 0.684 — 0.769 — 0.846 — RD⁃GCSL 0.626 ↑1.8% 0.702 ↑1.8% 0.803 ↑3.4% 0.880 ↑3.4% Twitter16 UDGAT 0.594 — 0.723 — 0.820 — 0.871 — RD⁃GCSL 0.626 ↑3.2% 0.743 ↑2.0% 0.838 ↑1.8% 0.889 ↑1.8% PHEME UDGAT 0.738 — 0.766 — 0.797 — 0.844 — RD⁃GCSL 0.745 ↑0.7% 0.776 ↑1.0% 0.807 ↑1.0% 0.856 ↑1.2%