Kernkonzepte
An LLM-based approach called HintDroid is proposed to automatically generate meaningful hint-text for text input components in mobile apps, helping visually impaired users better understand input requirements.
Zusammenfassung
This paper focuses on addressing the issue of missing or inadequate hint-text in text input components of mobile apps, which can significantly hinder visually impaired users from understanding and using the app's functionalities.
Key highlights:
- Motivational study finds that over 76% of text input components in popular mobile apps are missing hint-text, severely impacting accessibility for visually impaired users.
- Proposes HintDroid, an LLM-based approach that extracts GUI information, constructs prompts, and leverages in-context learning and feedback mechanisms to automatically generate meaningful hint-text.
- Extensive evaluation shows HintDroid outperforms 12 baselines, achieving over 82% boost in exact match compared to the best baseline. User study confirms its usefulness in assisting visually impaired users.
- Aims to invoke the community's attention to maintaining app accessibility, especially for text input components, and provides an automated solution to address the prevalent issue of missing hint-text.
Statistiken
The app name is "Flight", it has following activities: "Main, OneWay, RoundTrip, ..."
The current GUI page is "RoundTrip", it has following components:"Depart, Arrive, Departure time, ...", the upper part of the page is "Flight Search, ...", the lower part is "Departure time, ... ".
The text input of this page is "Departure time", its nearby components are "Flight Search, ... ".
Zitate
"It's helpful to show text in the element itself, in addition to making this hint-text available to screen readers."
"At least 2.2 billion people have near or distant vision impairment. In at least 1 billion of these, vision impairment could have been prevented or is yet to be addressed."