Retrieving memory as prompts for continual relation extraction

Open Webpage

Yini Zhang, Huanchun Peng, Yuxuan Zhang, Yuanxiang Li, Lei Huang

Published: 01 Dec 2024, Last Modified: 07 Nov 2025Expert Systems with ApplicationsEveryoneRevisionsCC BY-SA 4.0
Abstract: Highlights•A Memory Retrieval Enhanced Model(MREM) is proposed for CRE.•Memory samples are retrieved as prompts in MREM to reduce the risk of overfitting.•MREM fuses the knowledge of prompts and semantics to tackle catastrophic forgetting.•MREM shows outstanding performance on publicly available datasets.