Improving long-tail relation extraction via adaptive adjustment and causal inference

Jingyao Tang, Lishuang Li, Hongbin Lu, Beibei Zhang, Haiming Wu

Abstract: Highlights•We propose a model integrating adaptive adjustment and causal inference.•The adaptive adjustment module adjusts class margins and improves the generalization.•The causal inference module eliminates spurious correlations and enhances the robustness.•Our model achieves state-of-the-art performance on the NYT24 and NYT datasets.