An Approach to Ambiguity Analysis in Safety-related Standards Isabella Biscoglio, Alessandro Coco, Mario Fusani, Stefania Gnesi, Gianluca Trentanni Systems & Software Evaluation Center and Formal Methods Laboratory, ISTI-CNR Pisa, Italy ssec@isti.cnr.it Abstract— Standards for systems and software lifecycle processes have become rather popular in the last decade. Being expressed in natural language, their requirements, or clauses, are exposed to the risk of ambiguity, vagueness and subjectivity, even when safety of people and environment is the Standard’s main concern. The paper addresses some issues of this problem and presents an experimental approach to the determination and evaluation of a set of properties of the clauses, which capture the notion of the quality of their expressions. The approach adopts a rather intuitive quality model for the English language and includes the use of a tool for sentence processing. Results of a descriptive analysis of some well-known, safety-related Standards for different software application domains are shown and discussed. Keywords: Quality; standards; standardization; natural language analysis I. INTRODUCTION The role of Standards in software/systems engineering has been discussed for years in various environments and in various ways. The eventuality of their adoption as reference guidelines or strict, law-enforced prescriptions has hardly been welcome in the developers’ community, as they tend to be considered a restraint to creativity and to technological evolution. Yet, large manufacturing companies and governmental authorities, with different reasons and merits that will not be discussed here, often refer to Standards as recommended or mandatory process requirements for their suppliers. Differently from functional Standards, where measurable input-output relationships are expressed as product requirements, these Standards are mostly concerned with the task of ascertaining the existence and the efficacy of defined engineering and management practices. So, the impact of a software-related product into its operating environment may well depend on those Standards and, more particularly, on the related activities of: • Standards making (the main actors being standardization bodies); • Standards adoption (the main actors being suppliers’ and users’ organisms); • Standards conformity verification (the main actors being independent V&V organisms, such as certification bodies). Although these dependencies have been addressed in literature in various ways [1, 2, 3, 4, 5], little or no investigation has been reported so far on the role that the quality of the expressions used by the Standards plays there. The authors consider this as a rather important aspect, in that, being decided upstream in all the mentioned activities, the way Standards clauses are constructed can sensibly influence the way they are used and the related conformance verified, with evident impact of conformant-declared products into the users’ environment. The purpose of this paper is to present an experimental approach to the evaluation of some existing Standards’ requirements against defined quality attributes. The approach is based on a rather intuitive Quality Model for the natural language used in Standards and includes the use of a tool, called QuARS, for analyzing the sentences in the Standards requirements. QuARS (Quality Analyzer for Requirements Specification), conceived for analyzing software requirements [12, 13, 14], performs, according to a specified quality model, an initial parsing of natural language (NL) requirements for the automatic detection of potential linguistic defects that can determine ambiguity problems, impacting the subsequent stages of development. Its underlying process identifies both candidate defective words, by matching them with a corresponding set of dictionaries, as well as potential ambiguous sentence structures, by exploiting a syntactic parser. The process is split in two parts: (i) the "lexical analysis" capturing optionality, subjectivity, vagueness and weakness defects, and (ii) the "syntactical analysis" capturing implicitly, multiplicity and under-specification defects. The authors are aware, being involved in Standards developing groups themselves, that there are strict rules on the use of terms and expressions such as “should”, “shall”, “must”, and take for granted that they are carefully adopted and checked. What was investigated in the work reported by this paper goes beyond these basic rules to include syntactic and lexical aspects that may affect the users’ ability to understand a Standard requirement. 2010 Seventh International Conference on the Quality of Information and Communications Technology 978-0-7695-4241-6/10 $26.00 © 2010 IEEE DOI 10.1109/QUATIC.2010.83 461