知识融合研究方法

Review of Knowledge Fusion Research Methods

知识融合是知识管理的一个重要部分,它可以融合、转换分布式的信息源,是一种将多个信息源的不同知识进行合并,从而产生新知识的过程。知识融合能对不同的知识库进行深度挖掘,实现不同知识系统资源之间的共享,从而获取有价值或可用的新知识。该文主要对知识融合的相关研究方法进行了综述。首先,对国内外研究现状进行了归纳整理并对语义规则、贝叶斯网络等知识融合算法进行了总结,对所用算法的目的和未来研究方向进行了描述。其次,重点从知识融合定义、模式、框架三个方面进行综述,意在解释当前知识融合由于多学科领域交叉,概念边界模糊,没有形成统一的框架的问题,为研究知识融合的通用框架提供了新思路。再次,详细讨论了机器学习方法、深度学习方法等知识融合的前沿方法,并结合具体的应用实例对知识融合方法的特点进行了分析。最后对知识融合领域应用状况及现阶段存在的问题和发展方向进行了总结和展望。

Knowledge fusion is an important part of knowledge management.It can integrate and transform distributed information sources,which is a process of merging different knowledge of multiple information sources to produce new knowledge.Knowledge fusion can conduct deep mining of different knowledge bases and realize the sharing between different knowledge system resources,so as to acquire valuable or available new knowledge.In this study,we mainly review the relevant research methods of knowledge fusion.Firstly,the research situation at home and broad is summarized,knowledge fusion algorithms such as semantic rules and Bayesian networks are summed up,and the purpose and future research direction is described.Secondly,the focus is on the three aspects of the definition,model and framework of knowledge integration,aiming to explain the current knowledge integration because of interdisciplinary intersection,vague conceptual boundaries and no unified framework,which provides new ideas for the general framework of knowledge integration.Thirdly,the cutting-edge knowledge fusion methods of machine learning methods and deep learning methods are discussed in detail,and the characteristics of knowledge fusion methods combined with specific application examples are analyzed.Finally,the application status and existing problems and development direction of knowledge integration are summarized and predicted.