TTM-RE：记忆增强的文档级关系抽取
TTM-RE: Memory-Augmented Document-Level Relation Extraction
AI总结：
本文提出TTM-RE方法，通过可训练记忆模块和噪声鲁棒损失函数，有效利用大规模噪声训练数据，在ReDocRED等文档级关系抽取基准上取得最先进性能。
摘要：
文档级关系抽取旨在对文档内任意两个实体之间的关联进行分类。我们发现，先前用于文档级关系抽取的方法在充分利用大量具有不同噪声水平的训练数据方面效果不佳。例如，在ReDocRED基准数据集上，基于大规模、低质量的远程监督训练数据训练的最先进方法，通常并不比仅基于较小规模、高质量的人工标注训练数据训练的方法表现更好。为了充分释放大规模噪声训练数据在文档级关系抽取中的潜力，我们提出了TTM-RE，这是一种新方法，它将一个可训练的记忆模块（称为Token Turing Machine）与一个考虑正样本-无标注样本设置的噪声鲁棒损失函数相结合。在文档级关系抽取基准数据集ReDocRED上的大量实验表明，TTM-RE达到了最先进的性能（绝对F1分数提升超过3%）。消融研究进一步展示了TTM-RE在其他领域（生物医学领域的ChemDisGene数据集）以及高度无标注设置下的优越性。
英文摘要：
Document-level relation extraction aims to categorize the association between any two entities within a document. We find that previous methods for document-level relation extraction are ineffective in exploiting the full potential of large amounts of training data with varied noise levels. For example, in the ReDocRED benchmark dataset, state-of-the-art methods trained on the large-scale, lower-quality, distantly supervised training data generally do not perform better than those trained solely on the smaller, high-quality, human-annotated training data. To unlock the full potential of large-scale noisy training data for document-level relation extraction, we propose TTM-RE, a novel approach that integrates a trainable memory module, known as the Token Turing Machine, with a noisy-robust loss function that accounts for the positive-unlabeled setting. Extensive experiments on ReDocRED, a benchmark dataset for document-level relation extraction, reveal that TTM-RE achieves state-of-the-art performance (with an absolute F1 score improvement of over 3%). Ablation studies further illustrate the superiority of TTM-RE in other domains (the ChemDisGene dataset in the biomedical domain) and under highly unlabeled settings.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
