A myriad of resources can be found on the Web today, and finding (topically) relevant resources for a given information need is a daunting task. Even if relevant resources can be found, they may not be apt for the searcher in a given context: some properties of the resource may be “wrong” for his current context. Such issues can often be resolved by means of transformations. In this paper we discuss an algorithm for selecting candidate transformations for a given situation and present our first experiences with this algorithm.