Serial Contrastive Knowledge Distillation for Continual Few-shot Relation Extraction

TL;DR

This paper introduces SCKD, a novel model for continual few-shot relation extraction that addresses catastrophic forgetting and data sparsity through serial knowledge distillation and contrastive learning, demonstrating superior performance on benchmarks.

Contribution

The paper proposes SCKD, a new approach combining serial knowledge distillation and contrastive learning to improve continual few-shot relation extraction.

Findings

SCKD outperforms state-of-the-art models on benchmark datasets.

SCKD effectively mitigates catastrophic forgetting.

SCKD enhances knowledge transfer and memory utilization.

Abstract

Continual few-shot relation extraction (RE) aims to continuously train a model for new relations with few labeled training data, of which the major challenges are the catastrophic forgetting of old relations and the overfitting caused by data sparsity. In this paper, we propose a new model, namely SCKD, to accomplish the continual few-shot RE task. Specifically, we design serial knowledge distillation to preserve the prior knowledge from previous models and conduct contrastive learning with pseudo samples to keep the representations of samples in different relations sufficiently distinguishable. Our experiments on two benchmark datasets validate the effectiveness of SCKD for continual few-shot RE and its superiority in knowledge transfer and memory utilization over state-of-the-art models.