Evolving and maintaining software requires adequate documentation of its implementation. However, due to the software's constant evolution, the documentation and implementation do not remain synchronised. Intentional software views have been proposed as a documentation technique to alleviate this problem. Creating such views is not at all a trivial task, however. In this paper, we propose to use a learning algorithm that derives such intentional software views from extensional software views, which are much easier to build. The resulting approach combines the advantages of intentional software views with the ease of constructing extensional views.