Abstract: generation, conversions between data resource types, etcetera. We hypothesize that search systems can benefit a great deal from having a set of transformations available, especially from a searcher point of view. Consider the following motivating example: Example 4.1.1 While on the road, a searcher uses his Pda to search the web for information on his stock. The Pda is equipped with software for reading pdf files and html files only. Using his favorite search engine he finds a spreadsheet. In this case it would be very useful if the search engine automatically transforms the spreadsheet to one of the formats which he can access on his Pda. Without such transformation the data resource would be completely useless at this point. In this case, the transformation can be said to be value adding. Even more, a broker that is capable of executing transformations on demand is considered to be value adding, as long as it executes “useful” transformations. We will get back to this discussion later. Transformations can have an effect on certain properties of data resources. In the given example, the transformation would have an effect on the data resource type. Other examples of effects would be: changing the resolution of an image, the document length, etcetera. The usefulness of a transformation thus depends on the effects that it has and the properties that the searcher desires. The remainder of this chapter is organized as follows. We will start by presenting a formal framework for transformations in Section 4.2. With this framework we specify what transformations are and how they behave. After that we extend this framework with complex transformations (such as composed transformations, or transformations on instances of complex types; see also Section 3.4.4) in Section 4.3. After that we study the effects that transformations may have, as well as present the formal relation between transformations, resource space and resource base. This is done in Section 4.4. Next, in Section 4.5, we will present some practical aspects that are relevant when implementing a system using such transformations. This includes aspects such as learning the effects of transformations and automatic transformation selection. Last but not least, we will present some experiments with transformations in Chapter 5. 4.2 Transformations Let TR be a set of transformations. The semantics of a transformation specify what this transformation actually does. The semantics of a transformation is given by the function: