Mining candidate web services from legacy code

Mining services from existing software system is a relevant challenge to enable the development of service-oriented systems. Very often, although a service realizing a feature needed to develop a service-oriented system is not available, such a feature might have been implemented in existing software systems. This paper proposes an approach for the identification of services in existing legacy code. The approach combines the use of textual similarity between a free-text query and the text extracted from the legacy code with a structural matching between a goal, expressed as a WSDL document, and XML-based interfaces of features reverse engineered from the source code. A performance assessment has been performed on two open source projects.