The term goal refinement is used to capture both the decomposition of a goal into a set of subgoals and the specialization of a goal. The need for goal refinement (and thus for a corresponding component in the RM) arises as presentation goals are often formulated at a high level of abstraction. During goal refinement, the content of the final presentation will be determined. The Content Selection component assists in carrying out this task. In a concrete system, this module may appear as a retrieval and filter component which communicates with the application expert. The output of Goal Refinement and Content Selection is a set of communicative acts and a structural description of the relations that may hold between these acts. Communicative acts are an extension of speech acts  to multimedia communications (see also [7, 8]).
As soon as communicative acts have been worked out, it must be decided
which modalities and media shall be employed to convey them best.
For this task, the RM includes the Media Allocation component.
Note that there are different approaches for media allocation. One
approach is to map types of data or information onto types of media
(e.g., database entries onto tables versus maps). A more compositional
approach to media allocation is obtained when matching properties of
information (e.g. quantitative versus qualitative, fixed versus
varying) with properties of media (e.g., persistent or not; static
versus dynamic, 2D vs. 3D, having a geospatial dimension).
As the Media Allocation component assigns the particular
modality/medium to communicative acts, the output of the
component are now called media communicative acts.
The task of the Ordering component is to determine
in which order the selected content should be communicated to the
user. Such an ordering will be constrained by the semantic and pragmatic
relationships between the media communicative acts.
As the semantic and pragmatic relationships
eventually must be made recognizable to the presentation consumer, the
ordering of the media communicative acts will constrain the preceding
generation steps. While the term ordering is related with the term
linearization as used in the field of text generation,
it is important to note that multimedia
presentations do not necessarily follow linear structures as written
text and speech do.
Since there are many dependencies among choices, the components of the content layer are all connected to each other in order to enable negotiations among them. However, actual system implementations do often not handle some dependencies and thus do not allow for negotiations between all the tasks in the Content Layer.