Robert Stowe

Read more links pass the audit and still fail the reader

A long wooden pier reaching into flat gray water from an empty beach under a low overcast sky
Contextual Associations Between Webpage Elements for Web Accessibility: An Empirical Study
[Context] Screen reader users navigating webpages by element list often encounter accessible names such as "Read more" that are valid under the W3C Accessible Name and Description Computation specification but uninterpretable in isolation. The surrounding elements that would make these names meaningful exist in the page but are not linked to the target by any mechanism. No prior work has empirically studied how to select which surrounding elements are contextually relevant to a given target. [Objective] This registered report investigates whether human-perceived contextual associations between webpage elements can be recovered from the accessibility tree using link prediction, and whether the learned associations generalize across websites. [Method] We will construct a dataset of human-annotated contextual associations on 35 websites, stratified across the Tranco top-million list, with three independent annotators per page. Each page is represented as a graph derived from its accessibility tree, augmented with spatial and semantic features from the DOM and CSS. We compare four machine learning models (MLP, GCN, GAT, and SEAL) against two heuristic baselines under leave-one-site-out cross-validation with a pre-registered statistical framework, using Hit@K and MRR. [Results] We have conducted a five-site author-annotated pilot study to establish the pipelines and parameterize the power simulation, with pilot Hit@10 ranging from 0.16 to 0.85 across four learned models and 0.08 to 0.30 across two heuristic baselines. The final results will be reported after the planned experiments and analyses are completed. [Conclusion] The study contributes a human-annotated dataset of contextual associations on webpages, an empirical evaluation of link prediction for context selection on accessibility-tree graphs, and a cross-site generalization analysis.
arxiv.org

A registered report on arXiv, from Kishan Rakesh and Shiyi Wei, takes on a problem every screen reader user knows. A page has twelve links that all say Read more. Each is valid. Each has an accessible name. Pulled into a list, which is how many screen reader users navigate, they are twelve identical entries pointing at twelve different things, and the context that would tell them apart is in the surrounding text, with no formal link to it.

The authors plan to study 35 websites across the Tranco top million, with three annotators per page, and to train models that recover the association between a link and its context from the accessibility tree. The five-site pilot is promising: learned models reach a Hit@10 score as high as 0.85 against 0.30 for the best heuristic.

What I take from it is less about the models than about the problem. A page full of Read more links can pass every automated check and most manual ones. The names are present. The guidelines are met. And the page is still hard to use, because the accessible name says what the link is called, not what it is for.

I think this is the clearest argument I have seen that a passing audit is not the bar. The fix is small and it belongs in the component contract: a link component that requires a description when its visible text is generic, so that Read more about pricing is what the screen reader gets, every time, without an audit ever seeing it.

How many links on your product's busiest page say the same thing?

Photo source: https://photos.robertstowe.com/beach