人工智能培训

搜索

人工智能论文:将深度学习与LOGIC FUSION集成在一起以进行信息提取(Integrating Deep Learning with Logic Fusio

[复制链接]
77677 发表于 2019-12-9 14:34:33 | 显示全部楼层 |阅读模式
77677 2019-12-9 14:34:33 978 0 显示全部楼层
人工智能论文:将深度学习与LOGIC FUSION集成在一起以进行信息提取(Integrating Deep Learning with Logic Fusion for Information Extraction)信息提取(IE)旨在从输入文本中生成结构化信息,例如,命名实体识别和关系提取。已经通过功能工程或深度学习为IE提出了各种尝试,但是大多数尝试未能将任务本身固有的复杂关系联系起来,这被证明是至关重要的。例如,两个实体之间的关系高度依赖于它们的实体类型。这些依赖性可以被视为可以有效表达为逻辑规则的复杂约束。为了将这种逻辑推理能力与深度神经网络的学习能力相结合,我们建议将一阶逻辑形式的逻辑知识集成到深度学习系统中,以端到端的方式进行联合训练。该集成框架能够通过逻辑规则通过知识正则化增强神经输出,同时更新逻辑规则的权重以符合训练数据的特征。我们证明了所提出的模型在多个IE任务上的有效性和一般化。
Information extraction (IE) aims to produce structured information from aninput text, e.g., Named Entity Recognition and Relation Extraction.Variousattempts have been proposed for IE via feature engineering or deep learning.However, most of them fail to associate the complex relationships inherent inthe task itself, which has proven to be especially crucial.For example, therelation between 2 entities is highly dependent on their entity types.Thesedependencies can be regarded as complex constraints that can be efficientlyexpressed as logical rules.To combine such logic reasoning capabilities withlearning capabilities of deep neural networks, we propose to integrate logicalknowledge in the form of first-order logic into a deep learning system, whichcan be trained jointly in an end-to-end manner.The integrated framework isable to enhance neural outputs with knowledge regularization via logic rules,and at the same time update the weights of logic rules to comply with thecharacteristics of the training data.We demonstrate the effectiveness andgeneralization of the proposed model on multiple IE tasks.人工智能论文:将深度学习与LOGIC FUSION集成在一起以进行信息提取(Integrating Deep Learning with Logic Fusion for Information Extraction)
URL地址:https://arxiv.org/abs/1912.03041     ----pdf下载地址:https://arxiv.org/pdf/1912.03041    ----人工智能论文:将深度学习与LOGIC FUSION集成在一起以进行信息提取(Integrating Deep Learning with Logic Fusion for Information Extraction)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

77677当前离线
新手上路

查看:978 | 回复:0

快速回复 返回顶部 返回列表