Abstract:
In large-scaled software development projects, different software practitioners may deal with different pieces of software requirements depending on their perspectives or perceptions of their shared problems. Each of the users may define his/her requirements from his/her own point of view using different terminologies. However, a group of software practitioners often needs to interact, collaborate or trace requirements among the group in order to achieve common goals in their development process.
To resolve the semantic heterogeneity arising from requirements tracing, we have developed an alternative framework, called the Multiperspective Requirements Traceability (MUPRET). The framework enables automatic generation of traceability relationships of multiperspective requirements artifacts at a fine-grained level by using ontology for the course of an integrating system. In view of that, we designed a requirements ontology as a knowledge management mechanism to represent multiperspective requirements artifacts of an individual software practitioner. The requirements ontology interjects mutual understanding without restricting the freedom in expressing requirements differently. Ontology matching is applied to automate the generation of traceability links by taking two requirements ontologies and producing correspondences (i.e., equivalence, more general, less general, mismatch and overlapping) between the ontology concepts that semantically correspond to each other. As a result, the traceability relationships can be automatically generated when a match is found in the requirements ontologies.
As a proof-of-concept, three hospital information systems and two university information systems were tested. The results from these case studies demonstrated above an 80% precision rate and approximately a 75% recall rate for traceability relationships that were automatically generated by the MUPRET in comparison to the traceability relationships identified manually by users.