| Summary: | At the center of software reuse is the search and retrieval of software components from large software libraries. Recent research has illuminated a promising approach called multi-level filtering that breaks the problem up into a series of increasingly stringent filters that move along a continuum of high recall, low-precision syntactic techniques towards the more computationally expensive, high-precision semantic techniques. In multi-level filtering, syntactic matching is decomposed into two phases: profile filtering and signature matching. This thesis presents improvements to the resolution of syntactic profiles where the intent is to increase precision without a loss in recall during profile filtering. Large integer representation of profiles and profile lookup tables lead to an optimal time and space solution to profile representation. Finally, a new approach to signature matching is proposed that provides early pruning of the search space in an effort to cut down the time it takes to find valid signature maps. The resulting software is mature enough for future integration with the other elements of multi-level filtering as well as inclusion in a CASE tool such as CAPS
|
|---|
