Abstract
Modern web environments typically present navigation barriers for assistive technologies when web content lacks explicit programmatic semantics. A system can utilize an intermediating browser layer configured to execute a multimodal edge machine learning model entirely on a device of a user. The system may scan Document Object Model (DOM) structures to extract layout contexts and generate synthesized structural attributes. The generated attributes can be dynamically injected into an accessibility tree of a browser. Such an approach may enable real-time local remediation of interactive elements, embedded frames, tabular data, and visual media without communicating information to a remote server, thereby preserving data privacy while yielding a natively navigable web structure for assistive software.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Tarun, Gudipati, "Improving Web Accessibility Using On-Device Models", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11826