Voice-Input Text Correction Using Contextual Probability Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech-to-text conversion technologies often make mistakes, particularly when users speak with accents, in noisy environments, or use uncommon acronyms, requiring manual user input for corrections, which is inconvenient.
Innovation Solution
An electronic device analyzes user corrections to similar text portions, reevaluates voice input, and applies correction probability scores to automatically correct or suggest corrections based on context and similarity, enhancing the speech-to-text conversion process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If speech-to-text conversion is performed automatically, then conversion speed and convenience are improved, but accuracy deteriorates due to mistakes with accents, noisy environments, or uncommon acronyms
Solution Approach 1:
The system implements feedback by detecting user corrections to speech-to-text conversions and using these corrections to identify and correct similar errors throughout the document. The system analyzes correction patterns, compares corrected text against remaining uncorrected text, and automatically applies corrections where similarity thresholds are met, creating a closed-loop system that continuously improves accuracy based on user feedback.
Solution Approach 2:
The system performs preliminary actions by proactively identifying and correcting errors before the user manually corrects them. By analyzing the first user correction and automatically applying it to similar instances throughout the document, the system prevents repeated manual corrections and reduces the overall time users need to spend correcting errors.
2Measurement precision
If manual text correction is required for speech-to-text conversion errors, then conversion accuracy can be improved, but user convenience and time efficiency deteriorate
Solution Approach 1:
The system implements self-service by automatically detecting and correcting its own errors without requiring continuous user intervention. When a user corrects an error, the system autonomously analyzes the correction, identifies similar errors throughout the document, and automatically applies the correction pattern, making the system self-correcting and reducing the need for ongoing manual attention.
Solution Approach 2:
The system uses feedback from user corrections to automatically improve accuracy. By monitoring correction patterns and applying them systematically throughout the document, the system converts manual correction efforts into automated improvement mechanisms, reducing the overall time users need to spend on corrections.
3Productivity
If automated correction is applied to all text portions, then correction efficiency is improved, but false corrections increase due to lack of context understanding
Solution Approach 1:
The system applies local quality by treating each text portion individually based on its specific characteristics. Instead of uniformly correcting all instances, the system evaluates each potential correction location in context, comparing it against the correction pattern while considering local textual features, and only applies corrections where the similarity threshold is met, ensuring appropriate local化处理 of corrections.
Solution Approach 2:
The system uses parameter changes by implementing a similarity threshold parameter that controls when corrections are applied. By adjusting this parameter, the system balances between correcting all potential errors and avoiding false corrections, allowing flexible control over correction aggressiveness based on the specific context and error patterns detected.
Data Source
AI summary
Automated text correction using correction probability scores, in which a mobile device receives a correction of text converted from voice input and compares corrected text data to other portions of text data to assess similarity of the corrected text data to each of the other portions of text data. The mobile device also reevaluates the voice input with the text correction as an input and evaluates context of the text correction. The mobile device then determines a correction probability score based on the comparison of the corrected text data to the other portions of text data, the reevaluation of the voice input, and the evaluation of the context. After determining the correction probability score, the mobile device compares the determined correction probability score to ranges of correction probability scores associated with different correction actions, selects a correction action based on the comparison, and performs the selected correction action.