Automated Speech Recognition Correction with Iterative Hint Words
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated speech recognition (ASR) systems suffer from limited accuracy due to issues such as words sounding similar but having different meanings, difficult-to-pronounce words, mispronunciations, and noise, leading to decreased system performance and user frustration.
Innovation Solution
A computer-based process that successively submits words or phrases from the transcription, along with related words and phrases, as hint words to the ASR system until a more proper transcription is generated, thereby improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ASR systems process speech input directly without additional corrections, then the processing speed is fast, but the transcription accuracy is limited
Solution Approach 1:
The patent introduces an intermediary correction system that sits between the initial ASR transcription and the final output. This correction system uses hint words, contextual analysis, and iterative refinement to improve transcription accuracy without requiring fundamental changes to the underlying ASR system architecture.
Solution Approach 2:
The system performs preliminary actions by pre-processing the initial transcription to identify potential errors before final output. It uses techniques such as detecting homonyms, analyzing contextual relevance, and preparing correction candidates in advance, which allows for accurate corrections without adding significant processing time during actual use.
2Measurement precision
If ASR systems use hint words and iterative corrections to improve accuracy, then the transcription accuracy increases, but the processing time increases
Solution Approach 1:
The system applies partial correction actions by focusing only on the most likely error sources in the transcription rather than re-processing the entire input. It uses selective hint word submission and targeted error correction, performing just enough correction iterations to achieve acceptable accuracy without exhaustive processing.
Solution Approach 2:
The correction system uses self-service mechanisms by automatically detecting its own transcription errors and correcting them without external intervention. It employs self-correction loops where the system monitors its own output, identifies inconsistencies, and iteratively refines the transcription based on contextual clues and hint words.
3Measurement precision
If ASR systems correct transcriptions by submitting multiple hint words iteratively, then the accuracy improves, but the computational cost increases
Solution Approach 1:
The system performs preliminary analysis of the initial transcription to identify high-probability error locations before investing computational resources in correction. By pre-processing to detect homonyms, analyze contextual relevance, and prioritize correction candidates, the system minimizes the number of expensive iterative ASR calls needed.
Solution Approach 2:
The system changes parameters such as hint word selection, correction iteration depth, and error detection thresholds dynamically based on the complexity of the input and the confidence level of the initial transcription. This adaptive approach optimizes computational resource usage by applying intensive correction only when necessary.
Data Source
AI summary
Systems and methods are disclosed and described for correcting errors in ASR transcriptions. For an incorrect transcription, different words or phrases from the transcription, and/or related words or phrases, are submitted as hint words to the ASR system, and the voice query is submitted again, to determine new transcriptions. This process is repeated with different transcription terms, until a different and more proper transcription is generated. This increases the accuracy of ASR systems.