Linked Data Fusion Based on Similarity Calculation and K-Nearest Neighbor

Abstract

The development of semantic web technology supports the continuous development of linked data and its applications. In order to make effective use of the growing linked data on the web, multi-source data should be fused, which is a key step in multi-source large-scale data analysis and management. Currently, merging these data can be difficult, because various sources usually provide multiple conflict descriptions for entities in the same real world. To complete the fusion, we need to solve the problem of object conflict. This paper proposes a linked data fusion method based on similarity calculation and k-nearest neighbor. This method has two contributions. Firstly, a similarity calculation method of linked data is proposed, which can effectively integrate URI nodes and blank nodes in linked data; Secondly, a literal type node fusion strategy based on k-nearest neighbor classification method is proposed, which realizes the automation of fusion and has the independence of data source. The results show that compared with other methods, this method can improve the conciseness and consistency and precision by up to 12.9%, 30.6% and 12%.