How do we declare the underlying dataset for a page involving custom subsets or views of the data?
When integrating data from heterogenous sources it is sometimes necessary to synthesise page URIs non-algorithmically from the underlying data. Alternatively views of data may be required that follow a clustering or structure that does not have a simple 1:1 correspondence with underlying data URIs.
Create data about your description pages and include foaf:topic and foaf:primaryTopic properties to link the page to the resources that it describes. When rendering these pages obtain the data describing the page then bring in descriptions of each resource referenced with foaf:topic and foaf:primaryTopic to build the base dataset for the page.
The BBC programme pages include information on a primary topic supplemented with additional data about other related topics. The data included on each page may vary depending on factors other than the type of resource being described.
Most database driven pages on the Web involve more than one type of data and augment a base dataset with related information by using multiple database queries. In many cases it is possible for the publisher to anticipate these arrangements and describe them as linked data. This can reduce multiple queries to a single query describing the resource and any related resources. The resulting dataset can be passed directly to a templating system for rendering.
The topics associated with a page do not need to be closely related in the underlying data or even connected at all. The page description gathers together a group of resources according to the precise context specified by the publisher without reliance on particular relationships pre-existing in the data.
Changing the level of detail for classes of page or even of specified individual pages can be done simply by updating the description of those pages and allowing the templating system to work with the new dataset.
An additional benefit is that the page structure of the site can also be made queryable so it would be possible to discover which pages include information about a specific subject, thereby presenting the possibility of automatic cross-linking.