Tracing architecturally significant requirements: a decision-centric approach

This thesis describes a Decision-Centric traceability framework that supports software engineering activities such as architectural preservation, impact analysis, and visualization of design intent. We present a set of traceability patterns, derived from studying real-world architectural designs in high-assurance and high-performance systems. We further present a trace-retrieval approach that reverse engineers design decisions and their associated traceability links by training a classifier to recognize fragments of design decisions and then using the traceability patterns to reconstitute the decisions from their individual parts.