ASR Transcript Correction Validation Using Phonetic Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional media servers lack the technical capability to distinguish between valid and invalid user-provided corrections to automated speech recognition (ASR) transcripts, leading to the incorporation of inappropriate or nonsensical suggestions into captions.
Innovation Solution
A machine learning (ML) model trained using positive and negative examples to identify valid suggestions, incorporating features such as sound similarities, frequency of occurrence, and user consensus to determine the validity of corrections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If user-provided corrections are incorporated into ASR transcripts without validation, then caption generation speed is improved, but caption accuracy deteriorates due to incorporation of invalid suggestions
Solution Approach 1:
An ML model is introduced as an intermediary component between user suggestions and ASR transcript incorporation. The model evaluates suggested corrections by analyzing features such as phonetic similarity, edit distance, and contextual relevance, acting as a mediator that filters valid suggestions while blocking invalid ones, thus maintaining both speed and accuracy
Solution Approach 2:
The system implements a feedback mechanism where the ML model continuously evaluates user suggestions against the ASR transcript and provides acceptance or rejection decisions. This feedback loop ensures that only corrections meeting quality thresholds are incorporated, maintaining caption accuracy while enabling rapid processing of valid suggestions
2Manufacturing precision
If an ML model is implemented to validate suggestions, then suggestion accuracy is improved, but system complexity increases
Solution Approach 1:
The ML model utilizes multiple parameters and features (phonetic similarity, edit distance, contextual relevance, frequency of occurrence) to evaluate suggestions. By changing and analyzing multiple parameters simultaneously, the system achieves accurate validation without requiring overly complex architectural designs
Solution Approach 2:
The patent replaces manual validation mechanisms with an automated ML-based system. This substitution eliminates the need for complex human review processes while achieving comparable or superior validation accuracy through algorithmic analysis of linguistic and phonetic features
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An automated speech recognition (ASR) transcript of at least a portion of a media content is obtained from an ASR tool. Suggested words are received for corrected words of the ASR transcript of the media content. Features are obtained using at least the suggested words or the corrected words. The features include features relating to sound similarities between the suggested words and the corrected words. The features are input into a machine learning (ML) model to obtain a determination regarding a validity of the suggested words. Responsive to the suggested words constituting a valid suggestion, the suggested words are incorporated into the ASR transcript. At least a portion of the ASR transcript is transmitted to a user device in conjunction with at least a portion of the media content. eodcmaster eodcmaster