ML Accessibility Testing via Screen Reader Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for testing website accessibility for individuals with disabilities, such as those with low vision, are inefficient and resource-intensive, often requiring remote processing and lacking in accuracy, which can lead to non-compliance with ADA regulations.
Innovation Solution
A machine learning-based website accessibility testing tool that operates in conjunction with a screen reader, converting audio output into simulated user interactions to test webpage elements for compliance with ADA rules, using a local device to identify and record compliance issues without remote processing, thereby conserving resources and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current methods for testing website accessibility are used, then testing can be performed, but the process is inefficient and resource-intensive requiring remote processing
Solution Approach 1:
The patent introduces an intermediary machine learning model that acts as a bridge between the screen reader interactions and compliance determination. This local ML intermediary processes screen reader audio outputs and simulates user interactions without requiring remote server processing, thereby improving testing efficiency while reducing computational resource usage through local execution.
Solution Approach 2:
The system creates a virtual copy of user interactions by converting screen reader audio outputs into simulated keystrokes and mouse clicks. This copying approach allows the system to test accessibility without requiring actual human users or remote processing, enabling efficient local testing that reduces both time and computational resources.
2Measurement precision
If current accessibility testing methods are used, then testing can be performed, but accuracy is lacking leading to non-compliance with ADA regulations
Solution Approach 1:
The system performs self-service by automatically determining ADA compliance through local machine learning processing of screen reader interactions. The ML model autonomously analyzes the website's behavior in response to simulated user interactions and generates compliance determinations without requiring external expert review, thereby improving accuracy while managing complexity through automated local processing.
Solution Approach 2:
The system implements feedback mechanisms where the machine learning model continuously monitors screen reader audio outputs and browser responses, adjusting its analysis based on the website's actual behavior. This feedback loop enables the system to accurately determine compliance by comparing observed interactions against ADA requirements, improving measurement precision through iterative local analysis.
3Reliability
If remote processing is used for accessibility testing, then comprehensive analysis can be performed, but network resources are consumed and processing time increases
Solution Approach 1:
The patent segments the accessibility testing function into a portable machine learning model that can execute locally on the user's device. By dividing the testing system into a local ML component that handles analysis and determination, the system achieves comprehensive reliability through accurate local processing while eliminating network dependency and reducing processing time through immediate local execution.
Data Source
AI summary
An apparatus includes a memory and a hardware processor. The processor receives a voice signal associated with an element of a website. Navigating to the element includes performing a user interaction with a browser configured to display the website. In response to receiving the voice signal, the processor converts the voice signal to an input command that simulates the user interaction and executes the input command. The processor then monitors a behavior of the browser and applies a machine learning algorithm to the behavior to determine whether the website is compliant with a set of rules. The machine learning algorithm determines whether the website is compliant based at least in part on the voice signal and the browser behavior. In response to determining that the website is not compliant with the set of rules, the processor records a violation in an error log.