Continual relation extraction model based on memory sample selection and historical feature refinement

doi:10.1016/j.asoc.2025.114485

Continual relation extraction model based on memory sample selection and historical feature refinement

2025-12-20

0

PRE

AI

DOI：10.1016/j.asoc.2025.114485

Original

Original request for help

Save

Abstract

En 中文

• Memory samples and historical features are used to enhance the CRE model (CREMH). • A refinement and integration representation method is proposed for historical features. • CREMH alleviates catastrophic forgetting by enhancing the mass density of memory samples. • CREMH outperforms baselines in both overall and single-type scores on two datasets.

Continual relation extraction model based on memory sample selection and historical feature refinement

Abstract

Journal

Organization