AUTOMATED CONDITION DETECTION IN REQUIREMENTS ENGINEERING

Abstract In product development, it is of great importance that a complete, unambiguous, and, as far possible, contradiction-free target system defined. Requirements documents complex systems can contain several thousand individual requirements, derived in an interdisciplinary manner and written natural language by many different stakeholders. Hence, errors, the form contradictions, cannot be completely avoided these today they must corrected manually with high effort. This paper presents important building block for automated contradiction detection quality analysis requirements documents. We discuss necessary identification conditions extraction verbal expressions associated condition effect, respectively. applied analyzed processing methods based on grammatical versus machine learning models. The models have been to 1,861 real-world requirements. Both approaches generate promising results, accuracy partly over 98%. However, structured specification texts, model preferable due lower effort preprocessing better usability.