学科知识图谱的动态生成及其在资源智能组织中的应用

Dynamic Generation of Subject Knowledge Graph and Its Application in Resources Intelligent Organization

我国《关于推进教育新型基础设施建设构建高质量教育支撑体系的指导意见》,明确将学科知识图谱作为数字教育资源组织管理的智能工具。学科知识图谱作为认知智能技术,可有效将各类学科知识与资源建序优化,并以动态语义图式结构,映射学习者高阶认知活动。已有研究提供了资源组织应用视角下学科知识图谱构建生成的本体建模与关系抽取技术,但尚未关注基于跨模态实体语义计算的学科知识图谱动态补全,致使数字教育资源智能组织框架产生“信息茧房”。针对如何应用学科知识图谱破解数字教育资源智能组织中的“信息茧房”问题,通过系统建模与应用原型模拟分析,构建应用自然语言处理与跨模态张量分解模型,动态生成学科知识图谱的开放技术动态框架,形成学科知识图谱的动态生成、动态补全与质量评估三个阶段的关键技术与方法,进而基于学科语义张量分解模型,提出数字教育资源智能组织策略。研究发现,学科知识图谱动态补全是提高学科知识图谱建模质量的关键,基于学科语义张量的图谱补全方法引入惩罚函数、高阶奇异值分解与正交迭代的张量分解模型,消解了资源智能组织的“信息茧房”问题。未来相关研究,可基于学科知识图谱的动态生成分析学习者资源应用决策行为的深层认知机理,从学习行为语义关联分析、个性化精准教学决策、学科协同人才培养机制等方面,进一步拓展与深化。

The guidance of China’s new education infrastructure clearly takes the subject knowledge graph as an intelligent tool for the organization and management of digital education resources,and the subject knowledge graph as a cognitive intelligence technology,which can effectively sequence and optimize all kinds of subject knowledge and resources,and map learners’high-order cognitive activities with dynamic semantic schema structure.Existing studies have provided ontology modeling and relationship extraction technology for the construction and generation of discipline knowledge graph from the perspective of resource organization application,but they have not paid attention to the dynamic completion of discipline knowledge graph based on cross-modal entity semantic computing,resulting in the“information cocoonroom”of the intelligent organization framework of digital education resources.Aiming at how to solve the problem of“information cocoonroom”in the intelligent organization of digital education resources by using subject knowledge graph,through system modeling and application prototype simulation analysis,an open technology dynamic framework for dynamically generating subject knowledge graph by using natural language processing and cross-modal tensor decomposition model is constructed,and the key technologies and methods in three stages of dynamic generation,dynamic completion and quality evaluation of subject knowledge graph are formed,and based on the subject semantic tensor decomposition model,the intelligent organization strategy of digital education resources is proposed.It is found that the dynamic completion of subject knowledge graph is the key to improve the modeling quality of subject knowledge graph.The map completion method based on subject semantic tensor introduces the tensor decomposition model of penalty function,high-order singular value decomposition and orthogonal iteration to eliminate the“information cocoonroom”problem of resource intelligent organization.Future related research can analyze the deep cognitive mechanism of learners’resource application decision-making behavior based on the dynamic generation of discipline knowledge graph,and further expand and deepen from the aspects of semantic correlation analysis of learning behavior,personalized and accurate teaching decision-making,discipline collaborative talent training mechanism and so on.