Automatic extraction of requirements expressed in industrial standards : a way towards machine readable standards ?

TL;DR

This paper presents a semantic approach to automatically extract and process requirements from industrial standards, facilitating machine-readable standards through analysis, information extraction, and ontology creation.

Contribution

It introduces a novel semantic method combining manual and semi-automatic techniques for extracting and representing requirements from normative documents.

Findings

Effective extraction of requirement sentences using semantic annotation.

Creation of an enriched ontology for electrical appliance standards.

Comparison showing advantages over previous approaches.

Abstract

The project, under industrial funding, presented in this publication aims at the semantic analysis of a normative document describing requirements applicable to electrical appliances. The objective of the project is to build a semantic approach to extract and automatically process information related to the requirements contained in the standard. To this end, the project has been divided into three parts, covering the analysis of the requirements document, the extraction of relevant information and creation of the ontology and the comparison with other approaches. The first part of our work deals with the analysis of the requirements document under study. The study focuses on the specificity of the sentence structure, the use of particular words and vocabulary related to the representation of the requirements. The aim is to propose a representation facilitating the extraction of information, used in the second part of the study. In the second part, the extraction of relevant information is conducted in two ways: manual (the ontology being built by hand), semi-automatic (using semantic annotation software and natural language processing techniques). Whatever the method used, the aim of this extraction is to create the concept dictionary, then the ontology, enriched as the document is scanned and understood by the system. Once the relevant terms have been identified, the work focuses on identifying and representing the requirements, separating the textual writing from the information given in the tables. The automatic processing of requirements involves the extraction of sentences containing terms identified as relevant to a requirement. The identified requirement is then indexed and stored in a representation that can be used for query processing.