Towards Semantic Description of Explainable Machine Learning Workflows

Abstract

Machine learning (ML) has outperformed humans in many areas, and that is why it has been adopted by a large variety of applications, such as computer vision, speech recognition, and robotics. However, their functioning and the reason why they generated specific results usually are not clear to the user, being often considered as black-boxes. Explainable Artificial Intelligence (XAI) aims to make AI systems results better understandable to humans, enabling the optimization of the learning models and trust by users. Semantic Web Technologies (SWT) has been applied to ML models because they provide semantically interpretable tools and allow reasoning on knowledge bases that can help explain ML systems. Nevertheless, current solutions usually limit their explanations to the logic of the results, lacking the description or explanations of the other steps of the ML process, which can restrict the understanding experience of the user, making it difficult to identify in which step corrections and adjustments should take place. In this paper, we give an overview of XAI and SWT, and discuss the importance of providing a holistic solution, by means of an ontology. This challenge has to be addressed to improve understandability of the whole ML process and explanation process.