Application of Machine Learning in the Web of Linked Data

Linked Open Data (LOD) refers to guidelines for publishing and connecting structured data on the internet. Utilizing web technologies like HTTP, RDF, and URIs, Linked Data establishes entities across diverse domains and links them through categorized connections, thus forming a web of data readable by machines rather than humans. The LOD, often dubbed the Web of Linked Data, is an ever-expanding realm of information. Beyond mere data accumulation, the LOD methodology involves establishing connections between datasets. LODs and ontologies offer a universal solution that facilitates system interoperability, allowing for the sharing and utilizing shared information. However, since not all LODs employ the same ontologies, the use of diverse vocabularies and ontologies by organizations and communities across different fields to formalize entities and their relationships poses challenges to interoperability between different sets of LODs. When integrating LODs from various ontologies into a single entity, missing links may arise, leading to what we refer to as missing link scenarios. This paper examines multiple missing link scenarios primarily stemming from scattered ontologies across LODs. Subsequently, we propose feature- and graph-based methods for identifying missing links between LODs, significantly leveraging diverse ontologies. This research aims to provide a comprehensive review and introduce missing link management in LODs, which can facilitate the discovery of more valuable data by establishing connections with other datasets and enabling its more effective utilization through inference and semantic queries and rules.