Machine Learning Testing: Error, Fault, Or Failure? An Ontological Approach.

Software testing is an essential and expensive activity within the discipline of software development. Consequently, multiple quality standards were developed to improve the software testing process. Although these guidelines are effective for conventional software, they require modification to align with the realities of designing learning-based systems. Employing them without reorienting concepts may result in misinterpretations for individuals engaged in the development of these applications. The present study introduces an ontology, MLTont, that enhances the understanding of the tests performed and the results obtained in learning-based applications.