一种增量式本体模型与数据模式映射的图谱实例模型构建演化方法

Ontology-Schema Mapping Based Incremental Entity Model Construction and Evolution Approach of Knowledge Graph

在智慧城市领域中,随着信息化技术的不断深入,各信息系统产生的海量数据不断增长,这些多源异构数据之间的语义互通成为了城市智能应用开发需要解决的重要问题之一。构建知识图谱是解决数据语义互通的常用手段之一。在建立知识图谱本体模型后,图谱实例模型的构建演化就成为支撑基于图谱的各类应用的关键技术。为此,如何将不断更新的数据源中的知识实例尽可能自动化地扩充到知识图谱中,成为了图谱构建的首要问题。现有的一些知识实例生成工具对数据导入的支持力度不足,用户需要对源数据进行复杂的预处理,将其转化为符合平台支持的导入数据格式。这导致预处理工作量大,且不能迅速地应对数据不断更新增长的情况。由于智慧城市领域中信息系统所产生的数据多为结构化或半结构化数据,文中提出一种增量式本体模型与数据模式映射的图谱实例模型构建演化方法,面向结构化或半结构化数据生成实例,并随着数据的更新,实现图谱实例模型的增长与演化。文中方法结合机器推荐与人机协同交互设计,针对不同数据源的特征抽取知识并将其正确地映射到本体模型中的概念实体上,实现领域知识图谱实例模型的增量扩充;并通过实体对齐、关系补全等方法,支持实例模型的持续演化。文中方法在企业信息领域知识图谱的构建场景中得到了验证,通过机器推荐和不去重,实现了实例高效且准确的生成,其有效性也得到了证实。

In the field of smart city,with the deepening of information technology,many systems generate massive data.Semantic communication among these multi-source heterogeneous data has become one of the important problems to be solved in the deve-lopment of urban intelligent applications.Building knowledge graph is one of the common means to solve the semantic communication of data.After establishing ontology,the construction and evolution of graph entity model becomes the key technology to support various applications.Therefore,how to automatically extend the knowledge entities from constantly updated data sources becomes the primary problem of knowledge graph construction.Some existing knowledge entity generation tools cannot provide sufficient support for data import,and users need to carry out complex preprocessing of source data to convert it into the data format supported by the platform.As a result,the workload of preprocessing is heavy,and the data cannot be updated and increased rapidly.To deal with structured or semi-structured data,this paper proposes an ontology schema mapping-based incremental entity model construction and evolution approach of knowledge graph,which achieves the growth and evolution of instance model as data update.Based on the combination of machine recommendation and human-machine interaction,according to the characteristics of different data sources,the knowledge is extracted and correctly mapped to the concepts in the ontology model.The conti-nuous evolution of the entity model is supported by means of entity alignment and relationship complement.The approach is verified in the knowledge graph construction scenario of enterprise domain.By machine recommendation and prohibiting duplicate checking,efficient and accurate entity generation is realized,which proves the effectiveness of the approach.