Content modelling

Also sometimes called data modelling, even though that is a related but slightly different discipline.

Content modelling almost invariably involves some sort of data modelling.

A very interesting thread came up on Twitter in July 2021 where @eaton talks about the ‘three basic approaches’ to content modelling. Here’s my redestillation of it with my own comments:

  1. Platonic content. Starting without content, or high-level definitions of content, for instance interviews or events
    • the ontological approach, trying to describe the nature of the content
    • often beginning with a rabbit-hole of a conversation following what is and interview, really?
    • I find this approach often leads to too much specificity in the content model for actual content creators and editors
    • (and consequently too fine-grained interfaces)
  2. Organic content. Starting with a corpus of existing content and making some sort of system to deal with it all.
    • the navigating trade-offs approach
    • this approach can lead to sub-optimal content models if you want to make sure that all the different existing content types are met.
    • can be dangerously time-consuming if combined with a strict ontological framework
    • sometimes its worth finding a sound way of archiving the content, and take a fresh approach to managing the new stuff
  3. Visual design-first. Starting with visual artefacts and trying to model content based on what it looks like in a given setting.
    • if this is the only approach, it almost always leads to a total mess
    • very hard to ‘port’ the content to new outputs, devices and platforms

Side note: It is also worth investigating at what level we are modelling the content. Are we aiming for an organization to start using new tools to create and edit the content? Or are we modelling for a CMS at the end of a publishing flow?

I want to also suggest a fourth approach:

  1. Workflow first. Really get to know how content editors and creators work and how/if they think about their content as data.
    • this is the user experience research-approach
    • how can you optimise their workflow?
    • how do they think?
    • can you find a content model that lets editors get on with their work without too much intereference from the model or the user interface?

I find that I never take a single approach, almost all projects I work on require me to work with all four approaches:

  • Theres usually a bit of existing content.
  • It has often been modelled with the visual design-first approach.
  • It lets you start creating the content without know exactly what type of content it will be.
  • It needs some sort of ontology for the computer to understand it better
  • But being too focused on the ontological details makes us too narrow minded and wastes a lot of time (if it isn’t tested)
  • Testing the model in several visual designs.
  • Testing the model with a content editor (often they say it is way better than what they currently have, but I can sense that it is far from perfect, too much detail)
  • Going back to the drawing table, trying to find organic ways of representing the data being conceptually the same or similar to the ontological model, but not as restrictive.

I’m currently researching into content modelling with a (user-centric) workflow first, ontology later approach, letting an editor do their job (create text, move it around, chop it, flip it) first, then they can easily modify the text and add things like links and relationships to other documents.

Are there ways the editor/creator-human can define the content model as they go?

The content model starts out as just thing which can be anything, but for ease, lets call it a text document. The content ‘stub’ could then be ‘ontologically enhanced’ if and when needed. Thus creating a sort of organic content model.

This might create an anarchich mess of linked content with varying ontologies, but it might also be what we need for unfettered, liberated thinking about content.

It removes the question ‘is this possible?’ from an editor when creating or editing content. Yes, it should be possible. You, as the manager and editor and creator of the content should at least have the option to describe the content you want without having to talk to a content modelling expert first. Agency to the user.

It again can be queried against using something like SPARQL.

You’d end up with a graph of linked content instead of a hierarchical structure.

Relationships between things could be weak (for better or worse), but should have a robust way of keeping them in sync should that be your intention.

Semantic triples (RDF triples)