Improving Sentence-Level Relation Extraction through Curriculum Learning

TL;DR

This paper introduces a curriculum learning approach for sentence-level relation extraction, effectively handling noisy and difficult data, leading to state-of-the-art F1-scores on TACRED and Re-TACRED datasets.

Contribution

It proposes a novel curriculum learning framework that sorts data by difficulty to improve relation extraction performance.

Findings

Achieved F1-score of 75.0% on TACRED

Achieved F1-score of 91.4% on Re-TACRED

Outperforms previous state-of-the-art methods

Abstract

Sentence-level relation extraction mainly aims to classify the relation between two entities in a sentence. The sentence-level relation extraction corpus often contains data that are difficult for the model to infer or noise data. In this paper, we propose a curriculum learning-based relation extraction model that splits data by difficulty and utilizes them for learning. In the experiments with the representative sentence-level relation extraction datasets, TACRED and Re-TACRED, the proposed method obtained an F1-score of 75.0% and 91.4% respectively, which are the state-of-the-art performance.