Section: New Results

Collaborative Software Development Platforms

Participant : Isabelle Mirbel.

Today's Web has given rise to several platforms serving the purpose of collaborative software development. Thanks to these environments, it is possible, among others, for anyone to suggest new requirements for a software under development. A lot of requirements are thus proposed by users and it becomes difficult, after a while, for the persons in charge of the software which development is hosted by the platform to understand this large set of new requirements in its entirety. Therefore we proposed a tool to make large sets of requirements posted on collaborative software development platforms better workable despite the poor content of requirement body. Our aim was to propose an approach to automatically group similar requirements together in order to propose a limited number of requirement categories, thus improving the review process. As requirements expressed on collaborative software development platforms are usually very short and their content not very structured, we proposed to exploit relationships between stakeholders and already processed requirements to break the whole set of new requirements into meaningful categories. Our tool relies on Semantic Web languages and Formal Concept Analysis to provide a 3 steps data analysis process. The data is first extracted from the platform and translated into RDF, then stakeholders' past activities are analyzed to finally get stakeholder categories in order to improve the review of newly posted requirements.

According to the experiments that we conducted, we noticed some limitations in our approach. When the contributing stakeholders are newbies with no previous participation in any blueprint or bug and when there is no sufficient number of evaluated blueprints or bugs. To cope with this limitation, we plan to evaluate stakeholders reputation by looking at their activities on the whole collaborative software development platform (and not only the project under consideration). The results of this research have been published in [23] .

In today's software development methodologies, User Stories (US ) are mostly used as primary requirements artifacts. They are used to express requirements from a final user point of view and at a low abstraction basis using natural language. Over the years, several informal templates have been proposed by agile methods practitioners or academics to guide requirements gathering. Consequently, these templates are used in an ad-hoc manner, each modeler having idiosyncratic preferences. In this context, we performed a study of templates found in literature in order to propose a unified model. We also proposed an RDFS translation of this model in order to allow the annotation of user stories, thus providing search and reasoning capabilities to agile methods practitioners. The results of this research have been published in [56] .