Dr Mario Fusani
The Institute of Information Science and Technologies, Italian National Research Council (ISTI-CNR), Pisa, Italy.

Towards a Machine Learning based approach to evaluate the quality of Safety-Related Standards

Abstract:

Technical Standards are a common reference for verifying, comparing, and approving technical processes and products. Can the quality of such processes and products be influenced by the quality of the Standards themselves? This work aims to investigate whether various qualities that can be associated with regulatory requirements are just a sort of document decoration or an asset that, if weakened, may cause significant losses in the very objectives of the Standards. Currently, the work is moving a step forward from some previous stages where Standards qualities were first classified, then measured in some of their aspects. Now, we want to take advantage of the enormous amount of work the international teams of experts have done for many years. This important value is accumulated in a huge amount of comments, issued to correct and improve the draft editions of the Standards. This research starts by collecting the results of quality analyses (performed by means of Natural Language Processing tools) of a set of draft Standards requirements. Such results behave as “features” with which the comments of experts are associated and synthesized, with an ordinal-scale score, as data “labels”. The whole collection can then be viewed as a “supervised, semi-formal dataset”, to which various Machine Learning training algorithms are being selected to check their ability to generalize the scores for new datasets, including those derived from the final versions of the Standards.
The validation of quality analyzers, as well as automated aids for the standard makers, are among the issues that can be expected from this work.

Short bio:

Dr Mario Fusani is a Research Associate with the ISTI-CNR, an Institute of the Italian National Research Council in Pisa, Italy, where he has been working since 1973. In 1986 he founded the System and Software Evaluation Center in the same Institute, active since then. He has been working in various international working groups for standardization. His current research interest is in software-intensive system safety assurance and in safety-related standard comparison and evaluation.