【作者】 葛岩；

【导师】 马捷；

【作者基本信息】 吉林大学 ， 图书情报与档案管理， 2024， 博士

【摘要】 应急管理能力是国家治理能力的重要组成部分。新时代国家治理体系建设对应急管理的能力、体制和技术提出了新要求,从情报学研究视角,全面加强应急管理体系和能力建设,提升应急管理的智能化和精准化,实现面向知识组织的智慧服务,是亟待解决的重要议题。此外,根据文献梳理和实地调研,发现应急管理活动仍存在应急知识整合程度较低、知识组织体系概念颗粒度较粗以及应急管理部门快速协同效率低下等问题。突发事件的应急知识组织是实现应急管理信息化的基础,因此,面向突发事件的应急知识组织研究成为备受关注的议题。本文以应急知识表示与知识推理为研究对象,在系统梳理应急知识表示与知识推理相关文献的基础上,运用内容分析、案例分析和机器学习等方法,研究应急知识表示、知识推理和知识服务等问题,建立基于知识推理的应急知识库,以期为用户提供高质量的应急知识服务。本文主要工作和研究结论如下:第一,应急知识表示的理论模型构建。应急知识体系是描述特定专业知识综合的概念性术语,构建应急知识体系是开展应急知识表示的基础。首先,研究分析突发事件的内在要素及关系,分析在应急决策过程中,应急预案和应急案例等显性知识的重要性。其次,从政策工具视角,提取省级专项应急预案的知识结构要素;基于内容分析法,提取应急案例的知识构成要素。最后,实现面向应急预案与应急案例的应急知识概念级融合,构建基于知识元的应急知识表示的理论模型,提炼突发事件情景知识元、应急资源知识元、应急行动知识元和应急效果知识元,为后续应急知识表示与知识推理研究提供理论框架。第二,面向多层领域本体的应急知识表示。通过分析和研究本体建模的理论、技术和方法,构建面向多层领域本体的应急知识本体模型,即上层本体、领域本体和实例本体。首先,介绍ABC、SUMO和EVENT本体模型的核心概念及属性,分析其优缺点。其次,建立基于多层领域本体的应急知识表示框架,进行应急知识表示。详细分析应急知识的上层本体、应用本体和实例本体的构建流程,以实例本体为例,验证应急知识本体模型的有效性。最后,对应急知识本体进行定性与定量评价。第三,基于群体智慧的应急知识语料库构建。由多个标注者协作完成,通过多个标注者的交叉验证和互相纠错,实施协作平台的搭建、合理的任务分配和基于过程的质量控制,比传统的手工标注相比,基于群体智慧的应急知识语料协同标注方法的效率和准确性都提高。首先,考虑到需要标注四种类型的突发事件,因此,按照应急知识的上层本体结构,制定应急知识语料库的标注规范,确保标注标签的集中性和一致性。其次,提出基于群体智慧的应急知识语料协同标注模型,为后续应急知识语料库的标注行为提供理论指导。最后,采用数据标注平台Doccanno,对应急案例文本数据进行实体和关系标注,标注结果为后续实体识别、关系抽取和知识推理提供数据支撑。第四,融合语义解析与案例推理的应急知识推理。为了实现应急知识辅助应急决策,需要针对具体事件的应急处置方案,利用案例推理的方法,实现应急知识推理。首先,构建融合语义解析与案例推理的应急知识推理框架,阐述语义解析与案例推理之间的关系,分析基本的语义关系;其次,建立融合语义解析与案例推理的知识图谱推理模型,包括JSON数据语义解析、历史案例提取和案例相似度计算;然后,阐述本体与知识图谱之间的映射过程,构建应急知识图谱,实现应急知识的存储与查询,为后续第7章应急知识库的建立提供技术与数据支撑;最后,进行案例分析。选取典型应急案例事件,验证应急知识推理结果的可用性。第五,应急知识库构建。应急知识库是突发事件应急决策的重要依据,如何准确刻画用户信息需求是应急知识库构建的首要问题。首先,基于“手段—目的链”理论,通过内容编码、建立关联矩阵及绘制层级价值图,分析面向应急知识库的用户信息需求。其次,分析面向用户信息需求的应急知识库的构建框架与逻辑结构。然后,分析应急知识库的功能结构,实现应急知识库具体功能的设计与开发,阐述应急知识库的数据库设计与存储过程。最后,采用B/S网络结构模式,基于Thymeleaf+Springboot2.X+Neo4j开发框架,开发应急知识库原型系统,满足用户应急知识服务的需要。第六,提出应急知识服务策略。以提高应急知识服务水平为目的,从应急知识服务主体、服务客体和服务平台三要素,即政府、利益相关者和应急知识库平台,提出相应的应急知识服务策略。在理论层面,本研究丰富了知识组织理论,扩展了应急知识表示与知识抽取的方法研究,拓宽了知识库的应用领域。在实践层面,促进了应急管理工作的智能化与精准化,提升了政府应急管理能力和知识服务水平。更多还原

【Abstract】 Emergency management capability is a crucial component of national governance capability.The construction of the national governance system in the new era has put forward new requirements for the capabilities,institutions,and technologies of emergency management.From the perspective of information science research,how to comprehensively strengthen the emergency management systems and capabilities building,how to achieve the intelligence and precision of emergency management,and how to achieve knowledge-oriented smart services are urgent issues that need to be addressed.In addition,based on literature and field research,it has been discovered that emergency management activities still face problems such as low integration of emergency knowledge,coarse granularity of knowledge organization system concepts,and inefficient collaboration among emergency management departments.The organization of emergency knowledge organization is the foundation for realizing the informatization of emergency management.Therefore,research on the organization of emergency knowledge has become a highly concerned topic.This paper takes emergency text knowledge representation and organization as the research object.Based on a systematic review of relevant literature on emergency knowledge organization and knowledge reasoning,this paper uses content analysis,case analysis,and machine learning methods to study knowledge representation,knowledge reasoning and knowledge services,a knowledge-reasoning-oriented emergency knowledge base is established to provide high-quality emergency knowledge services for various users.The main conclusions of this paper are as follows:Firstly,the building of an emergency knowledge representation model.The emergency knowledge system is a conceptual term that describes the synthesis of emergency professional knowledge,and constructing an emergency knowledge system is the fundamental task for carrying out emergency knowledge representation.Initially,research and analyze the inherent elements and their relationships of emergencies,and the importance of explicit knowledge such as emergency plans and emergency cases in the emergency decision-making process was further clarified.Next,the provincial-level emergency plan for specific purpose was analyzed,from the perspective of the related policies.Based oncontent-analyzing method,the components of emergency cases were analyzed in this paper.Finally,in terms of emergency plans and emergency cases,a conceptual-level integration of emergency knowledge is what we would like to achieve.Based on the theory of knowledge elements,an emergency knowledge element model is constructed to extract knowledge elements such as emergency scenario knowledge elements,emergency resources,emergency actions,and emergency effects for emergency knowledge representation,providing a theoretical framework for subsequent research on emergency knowledge displaying and reasoning.Secondly,emergency knowledge representation based on multi-layer domain ontology.By analyzing and studying the theories,techniques,and methods of ontology modeling,we constructed an emergency knowledge ontology model for multi-layer domain ontologies,composed of upper level ontologies,domain ontologies and instance ontologies.Initially,the core concepts and attributes of ABC,SUMO and EVENT ontology models were introduced,the advantages and disadvantages were analyzed.Next,an emergency knowledge ontology model based on multi-layer domain ontologies was constructed for emergency domain knowledge representation.The construction process of the upper ontology,application ontology,and instance ontology of emergency knowledge was analyzed,and the instance ontology was taken as an example to verify the consistency and effectiveness of the multi-layer domain ontology model.Finally,the qualitative and quantitative evaluations of the emergency knowledge was performed.Thirdly,a corpus of the construction of an emergency knowledge for incidents to construted based on collective intelligence.The annotation method based on group intelligence significantly improved in annotation efficiency and quality compared to traditional manual annotation when facing a certain scale of corpus annotation tasks.This process not only requires fully utilizing the intelligence of each annotator,but also requires effective group collaboration,information discovery and intelligent aggregation.Initially,according to the ontology structure of emergency knowledge,establish annotation rules for emergency knowledge corpora corpus are established,laying a data foundation for the construction of emergency incident corpus;Next,a group intelligence based annotation model is proposed to provide theoretical support for the subsequent corpus annotation;Finally,using the data annotation platform Docanno,based on the ontology layer type names mentioned above,entity and relationship annotations were performed.The annotation results provide data support for subsequent entity recognition,relationship extraction,and knowledge inference.Fourthly,emergency knowledge reasoning that integrates semantic analysis and case-based reasoning.In order to achieve emergency decision-making and infer specific emergency response plans,this study proposes a knowledge graph based method to infer specific solutions.Initially,an emergency knowledge reasoning framework integrating semantic analysis and case-based reasoning was constructed,and the relationship between semantic analysis and case-based reasoning was elucidated.The basic semantic relationships were analyzed;Secondly,a knowledge graph inference model that integrates semantic analysis and case-based reasoning was established,including JSON data semantic analysis,historical case extraction,and case similarity calculation;Then,the mapping process between the ontology and the knowledge graph was established,and an emergency knowledge graph was constructed to store and query emergency knowledge,providing technical and data support for the establishment of the emergency knowledge base in Chapter7;Finally,case analysis was conducted.Typical emergency case events were selected to verify the usability of emergency knowledge reasoning results.Fifth,the construction of emergency knowledge base.The emergency knowledge base lays an important foundation for emergency decision-making,how to accurately characterize user information needs is the core issue of constructing an emergency knowledge base.Initially,based on the "means-end chain" theory,further analyze the user information requirements for emergency knowledge bases through content encoding,establishing correlation matrices,and drawing hierarchical value maps.Secondly,analyze the framework for constructing an emergency knowledge base for multi-agent needs,analyze the conceptual and logical structure design of the emergency knowledge base,and set the functional modules of the emergency knowledge base.The overall structure and function of the emergency knowledge base were designed based on user knowledge needs,on compessing both structured and unstructured storege processes;Finally,this system adopted the B/S network architecture mode and based on the Thymeleaf+Springboot2.X+Neo4j framework to develop an emergency knowledge base prototype system to meet the needs of emergency knowledge services.Sixthly,an emergency knowledge service strategy was proposed.With the aim of improving the level of emergency knowledge services,corresponding emergency knowledge service strategies were proposed from the three elements of emergency knowledge service subject,service object,and service platform,namely government,stakeholders,and emergency knowledge base platform.At the theoretical level,this study enriches the theory of knowledgeorganization,expands the research on methods of emergency knowledge representation and knowledge extraction,and broadens the application fields of knowledge bases.On a practical level,it promotes the intelligentization and precision of emergency management work,and enhances the government’s emergency management capabilities and service levels.更多还原