I'd have to point you to this BBC Internet Blog post for an excellent reference.
This blog post describes the technology strategy the BBC Future Media department is using to evolve from a relational content model and static publishing framework towards a fully dynamic semantic publishing (DSP) architecture.
You will need to define and describe a set of ontologies describing the structures and, importantly, relationships between content.
The hierarchical way to describe the structures is largely based on the premise of 'pages' as the core building blocks of the site. The thinking needs to be shifted to content items.
When you can get stakeholders thinking about content items combining dynamically to construct a page experience (i'll explain this shortly), you have half the battle won. This can be quite difficult.
One of the example issues you might face: In order to get Legal sign-off for a banking website we needed to submit highly accurate descriptions of all items that would appear on each 'page' month's before go-live. Including advertising and side-panels (yeah, I know...).
The problem is that the idea of a static page is no longer relevant. I referred to the current situation as 'page experiences'. They could be different combinations of content at different times for different people. Describing a page becomes a hazy task.
In summary:
You will need to define a model describing such things as classes/categories, sub-classes, relationships or attributes. The model is a logical conclusion of a content audit and strategy. [See BBC example below]
Remember: this is not final. The model is a living thing that will, and should change over time to reflect behaviour. What you are doing is setting the scene from what is currently known.
But how to keep the stakeholders happy? One thing I'd recommend is creating some scenarios to show how these models work to create page experience for particular personas at particular times.
Another method I used is to create a set of example wire-frames showing individual components (for content) plus a set stitching them together in representative patterns (templates with sample content). One these, I annotate the patterns of tags that determine each component. This helps make the link between your model and what a user might see.
BBC model example:

Building the model
I don't think I could describe this better than this quote from the BBC blog:
(RDF - Resource Description Framework - is based upon the idea of making statements about concepts/resources in the form of subject-predicate-object expressions. These expressions are known as triples in RDF terminology. The subject denotes the resource; and the predicate denotes traits or aspects of the resource and expresses a relationship between the subject and the object.
In some projects I've worked on we didn't need to go to this rigorous extent. Instead we had 3-4 sets of tag structures that not only described the content but the relationship that decided when they should appear - the content and marketing strategy.