Recovering semantic traceability between requirements and design for change impact analysis

Abstract

One of the ultimate challenges in change impact analysis is traceability modeling between several software artifacts in the software life cycle. This paper proposes a traceability approach that relates requirements and design artifacts modeled in UML. Our method faces two essential challenges: the semantic ambiguities of requirement artifacts that could be written in different natural languages and the heterogeneity of the artifacts that have to be traced (textual description, UML diagrams, etc.). To face these challenges, our method determines the semantic relationships between the requirements modeled with the use case diagram and design modeled with the class and the sequence diagrams through a semantic model which is an intelligent natural language processing technique that analyzes the semantics among the sentences, regardless of the language they are written with. Thanks to the semantic model our approach compares similarities between words having the same role, which makes it more efficient than computing similarities between words of different kinds. The empirical investigation demonstrates the advantages of the semantic traceability using a semantic model compared to the use of an information retrieval technique.