A Deep Learning-Based Framework for Handling Incompleteness and Detecting Errors in Linked Data Applied to the UniProt Dataset

Despite its potential for interlinking diverse datasets across domains, linked data is frequently hampered by in-complete information, incorrect relationships, and erroneous data entries. This paper presents a comprehensive framework designed to address the challenges of uncertainty and errors in linked data, specifically focusing on the UniProt dataset. The proposed framework integrates advanced deep learning models to tackle three key areas: missing type detection, missing link detection, and erroneous triple detection. The ultimate goal is to enhance the quality and reliability of the UniProt dataset, a vital resource in bioinformatics, by generating accurate triples and filtering out errors. This framework represents a significant step toward improving data completeness and integrity, thereby supporting more reliable research outcomes and applications.