【作者】 王宁；

【导师】 宋美娜；

【作者基本信息】 北京邮电大学 ， 计算机科学与技术（专业学位）， 2019， 硕士

【摘要】 领域知识图谱通常是从特定领域资源中抽取实体和实体之间的语义关系而构建的语义网络,它包含的知识体系具有很强的领域针对性和专业性。领域知识图谱构建平台则是为领域专家提供的,基于海量数据构建领域针对性强、准确度高的知识体系的简单易用的半自动化工具,应具备如下三个特点:构建流程定义完备;能够涵盖领域知识图谱构建过程中数据获取、信息抽取、知识融合、构建图谱、知识更新等各个流程;引入大数据处理能力;海量数据处理加工成为知识的过程离不开大数据平台的支持,因此平台需要具备大数据处理能力;简单易用,可操作性强;由于领域知识图谱具有很强的领域针对性和专业性,使用门槛过高不利于领域专家在构建过程中进行监督与干预。但是在当前大多公开的领域知识图谱构建平台中,还存在知识图谱构建流程定义不完善、缺乏大数据相关技术的支持和对于领域专家来说可操作性差的问题与挑战:当前大多公开的领域知识图谱构建平台对于知识图谱构建流程定义不完善,孤立地强调了知识图谱构建环节的某几个方面,诸如知识图谱中的数据采集、知识表示、图谱可视化等,不足以支撑全生命周期知识图谱构建工作;当前大多公开的领域知识图谱构建平台鲜少提及知识图谱构建过程中对应需要大数据相关技术的支持,缺乏对知识图谱实际构建过程的指导价值。在基于平台构建领域知识图谱的过程中,为保证精确度,往往需要领域专家的监督与干预,但是自然语言处理技术和大数据处理流程对于领域专家来说理解难度大,技术实现门槛高,可操作性差,对领域知识图谱的普及和应用产生了一定的限制。针对以上问题与挑战,本文重点围绕领域知识图谱的构建技术和流程进行研究与分析,完成了基于Web的领域知识图谱构建平台的设计与实现,主要研究内容有以下三项:1)设计并实现了基于Web的领域知识图谱构建平台,为领域专家提供构建流程定义完备、具备大数据处理能力且简单易用的知识图谱构建服务。在开发过程中为实现知识图谱构建流程的自定义编排,提出并实现了一种可视化Web服务组合编排技术。此外,还提出并实现了 DSACC(Dynamics Scheduling Algorithm for Concurrent Connections)算法,解决了知识图谱可视化过程中大数据量渲染的前端性能优化问题。2)提出并实现了一种基于大数据驱动的领域知识图谱构建方法,在完成第一项研究内容后,本文对知识图谱构建流程进一步总结,旨在研究在知识图谱构建过程中对应需要大数据相关技术的支持,为知识图谱的实际构建过程提供一定的参考价值。3)以基于Web的领域知识图谱构建平台为工具,以一种基于大数据驱动的领域知识图谱构建方法为指导,完成人工智能产业知识图谱的构建。图谱涵盖3458家人工智能企业,1087个人工智能领域技术标签,16324条专利数据,69866条相关新闻,全面展示人工智能产业发展现状,进一步证明平台与方法的有效性和完整性。更多还原

【Abstract】 The domain knowledge graph is a semantic network constructed by extracting the relationship between entities and entities from resources in a specific domain which has strong domain specificity and professionalism.The domain knowledge graph construction platform is a semi-automated tool for experts to build a highly targeted and accurate domain knowledge graph based on massive data.It should have the following three characteristics:The construction process is complete which covers the processes of data acquisition,information extraction,knowledge fusion,graph construction,knowledge update,etc.Processing massive data is inseparable from the support of big data platform,so the platform needs to have big data processing capability;Because the domain knowledge graph has strong domain specificity and professionalism,Excessive using thresholds are not conducive to the supervision and intervention of domain experts during the construction process.However,in the current public domain graph construction platfonn,there are many challenges in the definition of knowledge graph construction process,lack of support for big data related technologies and poor operability for domain experts.Most of the current domain knowledge graph construction platforms are incompletely defined for the construction process,and emphasize some aspects of the knowledge graph construction process in isolation,such as data collection,knowledge representation,graph visualization,which is unable to support the full life cycle knowledge graph construction;Most of the current domain knowledge graph construction platforms rarely mention the support of big data related technologies in the process of knowledge graph construction,which lack of guiding value in the actual construction process of knowledge graph.In the process of constructing the domain knowledge graph based on the platform,in order to ensure the accuracy,the supervision and intervention of the domain experts are often required,but the natural language processing technology and the big data processing flow are difficult for the domain experts to understand,the technical implementation threshold is high which imposes certain restrictions on the popularity and application of the domain knowledge graph.In view of the above problems and challenges,this paper focuses on the research and analysis of the domain knowledge graph construction technology and process,and completes the design and implementation of the Web-based domain knowledge graph construction platform.The main research contents are as follows:1)Design and implement a Web-based domain knowledge graph construction platform which provides domain experts with a knowledge graph construction service with complete definition process,big data processing capability and easy to use.In the development process,in order to realize the custom orchestration of the knowledge graph construction process,a visual Web service composition orchestration technology is proposed and implemented.In addition,the DSACC(Dynamic Scheduling Algorithm for Concurrent Connections)algorithm is proposed and implemented to solve the problem of front-end performance optimization of big data rendering in the process of knowledge graph visualization.2)A method based on big data driven domain knowledge graph is proposed and implemented.After completing the first part,this paper further summarizes the knowledge graph construction process,aiming to study the big data related technologies in the process of knowledge graph construction.The support of related technologies provides a certain reference value for the actual construction process of the knowledge graph.3)Using the Web-based domain knowledge graph construction platform as a tool,and based on a big data-driven domain knowledge graph construction method,complete the artificial intelligence industry knowledge graph construction.The graph covers 3,458 artificial intelligence enterprises,1,087 technology labels in the field of artificial intelligence,16,324 patent data,and 69,866 related news,which fully demonstrates the development status of the artificial intelligence industry and further proves the effectiveness and integrity of the platform and method.更多还原