科技情报对象关系抽取的技术选择  

Technology Choice of Relation Extraction of Scientific and Technical Objects

作　　者：刘秀磊[1] 王延飞[2] 刘思含 李红臣[3] LIU Xiulei;WANG Yanfei;LIU Sihan;LI Hongchen(Laboratory of Data Science and Information Analysis,Beijing Information Science and Technology University,Beijing 100101,China;Dept.of Information Management,Peking University,Beijing 100871,China;State Administration of Work Safety Communication Information Center,Beijing 100013,China)

机构地区：[1]北京信息科技大学数据科学与情报分析实验室,北京100101 [2]北京大学信息管理系,北京100871 [3]国家安全生产监督管理总局通信信息中心,北京100013

出　　处：《情报工程》2018年第3期39-47,共9页Technology Intelligence Engineering

基　　金：国家重点研发计划课题"基于云技术的煤矿典型动力灾害区域监控预警系统平台"(2016YFC0801407);国家自然科学基金"物联网搜索中异构本体的语义融合研究"(61601039);北京信息科技大学学校科研基金项目"本体匹配中词法分析关键技术研究"(1625008);北京信息科技大学软件工程专业学位点建设项目(5121723402);网络文化与数字传播北京市重点实验室开放课题"基于词法分析和语义分析的本体匹配研究"(ICDDXN006)

摘　　要：科技情报数据与日俱增,呈现海量、多源、异构的特性。针对上述特点,知识图谱能较深入地分析科技情报,实现对科技情报对象的感知和刻画。科技情报对象关系的抽取是知识图谱构建的关键步骤。本文在总结分析现有关系抽取技术和传统神经网络模型的基础上,提出一种基于长短期记忆神经网络和卷积神经网络深度机器学习的关系抽取技术,并通过实验证明其具有较好的性能,与其他运用神经网络的关系抽取技术相比,该技术的准确率、召回率和F值均有提高,运用该技术有利于增强对科技情报对象关系的感知刻画能力。The growing scientific and technical intelligence object data presented the characteristics of massive,multi-sourced and heterogeneous.For the above characteristics,knowledge graph can deeply analyze the scientific and technical intelligence and realize the perception,portrayal and merging of scientific and technical intelligence objects.Relation extraction of scientific and technical intelligence object is the key step in the construction of knowledge map.This paper proposes a relation extraction technology based on the long short-term memory and convolutional neural network deep machine learning.The results of experiment indicate that the new method obtain better performances compared with other neural networks extracted from relation including the accuracy,recall rate and F value.This new technology can help to enhance the portrayal and integration of science and te chnology intelligence objects.

关 键 词：关系抽取 科技情报 长短期记忆神经网络 卷积神经网络 

分 类 号：G35[文化科学—情报学]