Object-oriented Analysis Using Natural Language Processing
Richard George Dewar, Kangtong Li, Robert John Pooley · 2006
There is often a wealth of extant domain-specific, natural-language data available to help guide developers of object-oriented systems. This data is generally amenable to natural language processing in order to derive valuable design information. However, from a review of the field, we contend that current approaches have not been able to extract all the semantic and design detail that is present in such data. For instance: we see a lack of dynamic system representations; we detect reluctance for researchers to adopt hybrid solutions where users confirm and elaborate automated analyses; and we suggest there is work needed to determine a comprehensive repertoire of potential relationships between system components. In partial pursuit of such a manifesto, this paper mentions briefly our algorithmic proposal for processing natural language into UML with supplemental user involvement.