Recovering Legacy Requirements

It is common for organisations to introduce substantial changes to their structure and operations in order to adapt to new business environments. This often confers legacy status on their software systems because they can’t adequately support the new business processes. In this paper, we argue that it is necessary to recover the requirements of in-service legacy software to ensure that its evolution or replacement is properly informed by an understanding of what is redundant, what must be retained and what can be reused. Much of this information is often contained in documents. However, retrieval of the information is often difficult due to problems of completeness, quality and sheer volume. In the REVERE project we are integrating a number of techniques to provide a set of tools to help requirements engineers explore the documentation and reconstruct conceptual models of the software and business processes. At the core of this work is the exploitation of probabilistic NLP tools to provide a ‘quick way in’ to large, complex and imperfectly structured documents, saving much painstaking and error-prone manual effort.