Synthesized Speech Word Replacement for Reader Accessibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer reader tools struggle with generating synthesized speech that is difficult for users to understand, particularly due to phonetically similar words and lack of context, necessitating time-consuming manual testing.
Innovation Solution
A neural network-based system that automatically identifies words exceeding an understanding threshold by analyzing confidence scores and replaces them with easier-to-understand synonyms, enhancing accessibility testing and improving synthesized speech clarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If synthesized speech is generated from text data, then accessibility tools can provide speech output, but the synthesized speech becomes difficult to understand due to phonetically similar words and lack of context
Solution Approach 1:
The system uses a neural network to evaluate the synthesized speech and provides feedback by identifying phonetically similar words that cause confusion. This feedback loop allows the system to detect understanding difficulties and trigger replacement of problematic words with clearer alternatives, thereby improving the ease of understanding while maintaining pronunciation accuracy.
Solution Approach 2:
The system changes the parameters of the synthesized speech by replacing specific words with alternative words that have different phonetic characteristics. This parameter change in the speech output transforms difficult-to-understand phonetic patterns into clearer, more distinguishable words, resolving the contradiction between ease of understanding and pronunciation accuracy.
2Measurement precision
If manual testing of synthesized speech accessibility is performed, then accessibility can be evaluated, but the process becomes time-consuming
Solution Approach 1:
The system performs self-testing by using a neural network to automatically evaluate the synthesized speech for accessibility issues. Instead of requiring manual testing, the system independently analyzes its own output, identifies phonetically similar words, and generates replacement suggestions. This self-service capability maintains measurement precision while eliminating the time-consuming manual testing process.
Solution Approach 2:
The patent replaces the mechanical process of manual listening and evaluation with an automated neural network system. The neural network processes synthesized speech, identifies accessibility issues, and generates recommendations, substituting human manual testing with an automated intelligent system that achieves the same measurement precision much faster.
3Ease of operation
If words are replaced to improve understanding, then accessibility is enhanced, but the complexity of the speech generation process increases
Solution Approach 1:
The neural network acts as an intermediary component between the text-to-speech system and the final speech output. It processes the synthesized speech, identifies phonetically similar words, and generates replacement suggestions. This intermediary layer adds the necessary complexity only where needed to improve understanding, while keeping the overall system architecture modular and manageable.
Data Source
AI summary
A method, computer program, and computer system are provided for improving accessibility to computer reader tools. Data corresponding to one or more words to be displayed to a user is received. The received data is converted into synthesized speech. One or more words exceeding an understanding threshold value are identified from the synthesized speech. The understanding threshold value corresponds to a probability of difficulty associated with understanding the one or more words. One or more replacement words are retrieved for the one or more words exceeding the understanding threshold value. The synthesized speech is updated with the one or more replacement words.