Grant: Discovery Grants Program - Individual

Building high quality applications require an efficient access to specific knowledge. It is the case especially for software design and coding, where, software practitioners need to make decisions; they clearly need support for deciding what alternative is better for a specific context, and for identifying the effect of their decision on the quality of the software product. In recent work on software quality, we have used machine-learning techniques to capture knowledge from data (e.g., fault prevention knowledge, impact knowledge, etc.). However, in general we have to consider informal (or semi-formal) knowledge (texts, diagrams, statements, observations, behaviors, etc.) and then we have to structure it in order to finally formalize it and make it available for an automatic processing. A knowledge acquisition framework has to support this variety of sources; the ultimate goal of such a framework is to build knowledge bases that will support decision making for future projects. Considering knowledge implied in the whole development process is a vast and complex challenge. Thus, the objective of this proposal is to go towards an automated solution for manipulating knowledge for (only) software design and coding. We propose a knowledge-based architecture for decision making; it assumes (i) a knowledge representation language based on objects -allowing at least to express constraints and rules, but also other formal representations-, and (ii) inference mechanisms for reasoning on this knowledge. On the other hand, this system will be built on top of an existing framework for object-oriented applications analysis. Such an architecture will allow (1) software designers/programmers exploit knowledge bases during software construction and help them taking ""right"" decisions during design/coding. It will also allow (2) more experienced software designers/programmers manage the knowledge they have learned over their practice, and store it in formal and reusable schemes. Finally, it will help (3) capture and acquire knowledge from past software development projects, by analyzing their data with appropriate techniques.