Readability checks

[thibaudcolas]

Readability is fundamental to accessibility. One of the ways to improve text content is to have a clear target for reading level / reading age, which can be assessed with wagtail-readinglevel as a score displayed in rich text fields.

[lb-]

A quick dump of some of the tooling I have looked into for some similar checks for a work project.

• https://vale.sh/docs/vale-cli/installation/ -> Vale is quite a solid tool, we have also looked at it for potentially automated Wagtail documentation checks. It's quite configurable, written in Go but has compiled packages for npm and pypi as a CLI interface. It would not be able to be run client-side only though. I do like the way they approach config, it's usage agnostic yml files (which, in theory, could be dynamically generated).
• https://github.com/retextjs/retext -> retext is part of a suite of tools (includes Prettier), all JS, quite a powerful ecosystem that includes lots of different checks across different languages.
• https://github.com/textlint/textlint - textlint a similar set of tools for language / text checks but implemented in a very flexible way. The library and ecosystem is going through a bit of change though and some packages have not been updated in many years.

[thibaudcolas]

Moving to Future. We’ve made progress on new designs, and decided to take the concept well beyond readability checks. Readability checks remain a key element of this but the capabilities need to be more generic. For further details, see Content quality checkers #11063.

[thibaudcolas]

With Content metrics in page editor - word count and reading time #12020, we’re close to having the mechanism by which we can analyse whole-page content to produce metrics like readability. Once that’s in place, adding new metrics will largely be a matter of:

1. Determining where and how to place them in the UI.
2. Deciding on a formula that works for the breadth of CMS-managed content out there.

---

We have quite a solid plan for the UI, but not for the readability metric calculation / formulas. We’ll need to go through available readability formulas, and I’d recommend we pick one that:

1. Works well enough across all languages – or if not, can be configured differently per language for better results.
2. Doesn’t require a heavy JS package to implement.

Good references:

• Readable - Readability formulas
• Siteimprove - Readability tests

To progress with this, I’d recommend to:

• Review state of the art research papers about readability formulas, similarly to what @albinazs did for reading time (see my reading time notes). Ideally we’d find one or two meta-analysis reviewing results across multiple languages, published in journals with a decent impact factor, from the last 10 years or so.
• Review the formulas to find the ones that can work with metrics available in vanilla JS (or close to it). In particular things that are doable with the Intl.Segmenter API, or similar APIs, or even future proposals.

[streats]

I've started some research on this 🙂

[thibaudcolas]

Linking to Annie’s research and my comments for future ref: https://github.com/wagtail/roadmap/pull/79.