Intelligent support for retrieval and synthesis of patterns for object-oriented design

Abstract

Several decades of software engineering research confirm that effective reuse is the only realistic approach to meeting the ever-increasing demands on the software industry. Over the last few years, object-oriented paradigm has emerged as the natural foundational technology for reuse approaches, leading to the development of a number of reusable artifacts at different levels of abstraction and granularity. This research focuses on artifacts at one level: ‘patterns’. Patterns are groups of objects with stereotypical properties and responsibilities that can be applied by analogy to different domains. In this paper, we present a methodology for automating design of object-oriented systems based on intelligent retrieval and synthesis of reusable patterns. The methodology itself has an object-oriented flavor. It uses a set of techniques and rules aided by heuristics from natural language processing, automated reasoning and learning that are activated, as needed. Effectiveness of the methodology is evaluated using measures such as recall, precision, coverage and spuriousness.