Even word / office documents have structure, well kinda. They can be fragmented into chapters, sections… etc. Using semantic structures we can design very specific structures which give us good control over the authoring process.
For example in my name structure I can define the first- and surname obligatory and the middle name optional:
An (XML) editor will guide the author to respect these rules defined by the information architect. The outcome is consistent, controlled content.
It depends on the nature of the content if we need this control and also how deep structure is necessary. Dictionaries are probably the best example. They have complex content, usually written by many individuals in collaboration. In order to get consistent result, the authoring has to be controlled by a deep semantic structure, like this:
Here the dict element has more than one entries. An entry has exactly one form-group and one or more sense-group… etc.
In the office document world usually each document is individually formatted. Here we separate the content from the layout and place the structure in between them. Then the documents (content) and layout (styles) are both connected to the structure:
Please note the cardinality on the illustration above. It’s only worth to define the structure if we have many occurrences of the same document type. WIth other words if many documents share the same structure. No direct binding between the content and the layout, instead the content is controlled by the structure and the styles are bound to the structure (to XML elements). While we have a single structure (more about this later) we can have several styles, for example one for each publishing platform.