Characterising noise in ontologies

Abstract

This thesis presents a comprehensive investigation into the nature and impact of noise on on tologies and the response of neuro-symbolic reasoners to noise. The project aims to develop a mechanism for introducing noise into ontologies, particularly focusing on the ABox, and evaluate the performance of existing neuro-symbolic reasoners under varying noise levels. Key compo nents of the problem include quantification of noise, development of an experimental framework, identification of reasoning tasks, and selection of appropriate evaluation metrics. The proposed approach introduces noise programmatically into the ABox, evaluates conventional and neuro symbolic reasoners on datasets with varying noise levels, and maintains proper version control for reproducibility. Expected outcomes include characterizing the impact of noise on ontologies, assessing the robustness of neuro-symbolic reasoning systems to noise, and validating the effec tiveness of the proposed approach. The findings of this study will provide valuable insights into ontology engineering and reasoning algorithms in noisy environments.