Attempt to Recommend LOD Link Candidates Using Literal Similarity

Abstract

　In recent years, various data held by government agencies and other organizations have been released as LOD, and their utilization has also been promoted. When converting data to LOD, it is necessary to check each piece of data in both the linked source and destination data sets and link each source data appropriately. In this process, manually searching for candidate link data is costly. In this study, the similarity of strings linked to the subject of RDF describing LOD is calculated based on Levenshtein distance, and the similarity is used to select candidate linked data to be recommended as similarity of the subject. Because of the high processing cost of brute force processing, we propose to apply indexing using the feature vector method, and report the experimental results.