Automated Fine-Grained Requirements-to-Code Traceability Link Recovery

Problem: Existing approaches for requirements-to-code traceability link recovery rely on text retrieval to trace requirements to coarse-grained code documents (e.g., methods, files, classes, etc.), while suffering from low accuracy problems. Hypotheses: The salient information in most requirements is expressed as functional constraints, which can be automatically identified and categorized. Moreover, people use recognizable discourse patterns when describing them and developers use well-defined patterns for implementing them. Contributions: Recasting the requirements-to-code traceability link problem as an accurate matching between functional constraints and their implementation.