RH-Net: Improving Neural Relation Extraction via Reinforcement Learning and Hierarchical Relational Searching
论文下载:https://arxiv.org/abs/2010.14255
GitHub:https://github.com/wjn1996/PCNN_RL_HME
备注:论文一共更新了两个版本,论文和GitHub均提供的是最新的版本。
Abstract:
Distant supervision (DS) aims to generate large-scale heuristic labeling corpus, which is widely used for neural relation extraction currently. However, it heavily suffers from noisy labeling and long-tail distributions problem. Many advanced approaches usually separately address two problems, which ignore their mutual interactions. In this paper, we propose a novel framework named RH-Net, which utilizes \textbf{R}einforcement learning and Hierarchical relational searching module to improve relation extraction. We leverage reinforcement learning to instruct the model to select high-quality instances. We then propose the hierarchical relational searching module to share the semantics from correlative instances between data-rich and data-poor classes. During the iterative process, the two modules keep interacting to alleviate the noisy and long-tail problem simultaneously. Extensive experiments on widely used NYT data set clearly show that our method significant improvements over state-of-the-art baselines.
主题: 远程监督关系抽取、强化学习、噪声问题、非均衡长尾问题
动机:
-
远程监督方法以其启发式地与外部知识库进行对齐来自动化构建大规模数据集的方法在关系抽取任务中得以广泛使用,但其普遍遭遇到噪声和非均衡长尾问题;
-
噪声问题和非均衡长尾问题都归结于较强的启发式假设,使得大部分的句子被错误的标注。现如今大多数方法通过注意力机制或者强化学习等方法进一步降噪,但这也使得许多关系标签下有效句子数量大大降低,进一步加剧了长尾问题;
方法:
-
为了同时考虑到噪声和非均衡长尾两个问题,充分考虑到二者之间的影响交互关系,本文提出一种 RH-Net 方法,首先通过强化学习方法通过奖励机制促使模型自发地学习如何挑选高质量的句子,并将其喂入分层搜索模块中寻找正确的关系标签。
-
本文是首次将关系抽取任务转化为一种树的搜索问题。因此以FreeBase为例的知识图谱中关系标签是分层描述的,例如/people/person/placeOfBirth可以分层四层:root、/people、/people/person和/people/person/placeOfBirth。因此只需要让模型逐层进行搜索即可,这样便可以将具有相同子层的长尾关系和非长尾关系关联起来,通过非长尾关系对应的样本辅助预测长尾关系,达到提升关系抽取的效果。
插图:
论文正文:
模型架构图:
实验效果:
欢迎阅读与引用!!!