SPARQL-based linking of knowledge graphs

Abstract

Nowadays, both traditional and emerging enterprises rely heavily on Internet data, yet these data are disorganized and scattered in different data sources all over the world. In most of such data, the models may be different, but their contents may be similar, to utilize these data more efficiently, it is imperative to link these isolated data to create a unified view. Link discovery has always been a trending problem, and many companies tried to solve it. However, many of the solutions from these companies are not based on a common standard and thus result in a steep learning curve. This thesis focuses on the specification and development of link discovery proposal based on standard SPARQL, as well as, the implementation of several strings similarity metrics used for the linking. Specifically, it is a Resource Description Framework (RDF) linking task that produces connections between RDF resources from different or the same data sources, such as DBpedia, Wikidata and BNE datos. RDF linking is based on one or more link rules that specify the conditions that two RDF resources can be considered the same, such as text similarity or geographical position. SPARQL Protocol and RDF Query Language (SPARQL) is one of the most widely used and standard query language and protocol for Linked Open Data on the web or RDF triplestores. So, this TFM featured a SPARQL-based application to link knowledge graphs using Apache Jena as the SPARQL engine, a free and open-source Java framework for building Semantic Web and Linked Data applications. This TFM used multiple linking rules and evaluated the results with several data sets from the Ontology Alignment Evaluation Initiative (OAEI) competition in 2021 and achieved excellent results.