arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2008.12283cs.CL
收藏

实体和证据引导的DocRED关系抽取

Entity and Evidence Guided Relation Extraction for DocRED

JD AI Research(京东人工智能研究院);Stanford University(斯坦福大学)

Kevin Huang, Guangtao Wang, Tengyu Ma, Jing Huang

摘要:

文档级关系抽取是一项具有挑战性的任务,它需要对多个句子进行推理,以预测文档中的关系。本文提出了一个联合训练框架E2 GRE(Entity and Evidence Guided关系抽取)。首先,我们引入实体引导序列作为预训练语言模型(例如BERT,RoBERTa)的输入。这些实体引导的序列帮助预先训练的语言模型(LM)关注与实体相关的文档区域。其次,我们利用预训练语言模型的内部注意力概率作为证据预测的附加特征,引导预训练语言模型的微调,使其更关注实体和支持/证据句子。我们评估我们的E2 GRE方法DocRED,最近发布的大规模数据集关系抽取。我们的方法能够在所有指标的公共排行榜上取得最先进的结果,表明我们的E2 GRE在关系抽取和证据预测方面既有效又协同。

英文摘要:

Document-level relation extraction is a challenging task which requires reasoning over multiple sentences in order to predict relations in a document. In this paper, we pro-pose a joint training frameworkE2GRE(Entity and Evidence Guided Relation Extraction)for this task. First, we introduce entity-guided sequences as inputs to a pre-trained language model (e.g. BERT, RoBERTa). These entity-guided sequences help a pre-trained language model (LM) to focus on areas of the document related to the entity. Secondly, we guide the fine-tuning of the pre-trained language model by using its internal attention probabilities as additional features for evidence prediction.Our new approach encourages the pre-trained language model to focus on the entities and supporting/evidence sentences. We evaluate our E2GRE approach on DocRED, a recently released large-scale dataset for relation extraction. Our approach is able to achieve state-of-the-art results on the public leaderboard across all metrics, showing that our E2GRE is both effective and synergistic on relation extraction and evidence prediction.

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。

↑