Removing Biased Data in Knowledge Graph with Ontological Reasoning

Ritik Rawat, Ihsan Ibrahim, Takaaki Kawai, Naoki Fukuta

Abstract: In this paper, we propose an approach that will help reduce the biases in the Knowledge Graph (KG). While understanding new knowledge formed by connecting previous KGs, it is a difficult task to notice in the KG whether it is biased or not and in which direction it leads to the biased, since KGs represent complex relationships between entities, and biases may be introduced through these relationships. Overall, this study highlights the importance of considering the potential way to understand the knowledge in a fair way from the KGs. By realizing responsible KGs and conducting further research into the underlying mechanisms, we can work towards ensuring the implementation of strategies to reduce biases and create KGs with fair results.