Automatic Generation of Domain Representations Using Thesaurus Structures

Domain analysis was first used 15 years ago as one of the most important techniques for software reuse. Even today, new techniques appear every year, and different authors propose different domain representation structures to represent and store all the different software components and the relationships among them. These relationships among components are the kernel of the domain semantics. In this report, a set of techniques and tools is presented regarding mathematical, statistical, and neural fields that, when linked together, enable semiautomatically building domain representations and storing them in a thesaurus structure of software components. Thesaurus structures, widely used in information science, are presented as the domain-modeling key concept, due to their higher automation possibilities compared with previous structures. New metrics to evaluate the quality, consistency, and completeness of the domain model obtained through this technique are also presented.