Speech-to-Text Transcript Correction via Audio Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech-to-text conversion technologies face challenges in accuracy due to environmental noise, language dialects, regional accents, and individual pronunciation variations, leading to error-prone text generation that affects subsequent analyses.
Innovation Solution
A multi-threaded processor system performs combined pause detection and speaker diarization techniques to segment speech audio into more accurate units, using adaptive peak amplitude and connectionist temporal classification methods, and integrates acoustic and language models to generate transcripts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional speech-to-text conversion is used, then the process is simple and fast, but the accuracy is low due to environmental noise, dialects, accents, and pronunciation variations
Solution Approach 1:
The patent segments speech audio into multiple chunks and processes each chunk independently through multiple pause detection techniques. This segmentation allows the system to handle complex acoustic analysis in manageable portions, improving overall transcription accuracy while maintaining computational feasibility
Solution Approach 2:
The patent combines multiple pause detection techniques (peak amplitude detection, connectionist temporal classification, and sentence boundary detection) to identify sentence pauses more accurately. By merging multiple detection methods, the system overcomes the limitations of individual techniques and achieves higher transcription accuracy despite increased processing complexity
2Measurement precision
If speech audio is processed as continuous data, then processing is simpler, but sentence boundary detection and speaker identification are less accurate
Solution Approach 1:
The patent divides continuous speech audio into discrete chunks and processes each chunk through multiple detection techniques. This segmentation enables precise identification of sentence boundaries and speaker changes by analyzing acoustic features at chunk boundaries, thereby improving detection accuracy
Solution Approach 2:
The patent performs preliminary pause detection and speaker diarization on speech chunks before final transcript generation. By pre-identifying sentence boundaries and speaker segments, the system prepares structured data that improves the accuracy of subsequent transcription and analysis processes
3Measurement precision
If multiple pause detection techniques are used, then sentence identification accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent applies multiple pause detection techniques to segmented speech chunks rather than continuous audio. This segmentation reduces the computational burden of each individual detection operation, allowing multiple techniques to be applied efficiently while maintaining high pause detection accuracy
Solution Approach 2:
The patent implements a multi-stage detection process where different pause detection techniques are applied with varying levels of intensity. Not all detection techniques are applied with equal depth to every chunk, allowing the system to achieve sufficient accuracy while reducing overall processing time and computational resource consumption
Data Source
AI summary
A system, method, and computer-program product includes constructing a transcript correction training data corpus that includes a plurality of labeled audio transcription training data samples, wherein each of the plurality of labeled audio transcription training data samples includes: an incorrect audio transcription of a target piece of audio data; a correct audio transcription of the target piece of audio data; and a transcript correction identifier that, when applied to a model input that includes a likely incorrect audio transcript, defines a text-to-text transformation objective causing an audio transcript correction machine learning model to predict a corrected audio transcript based on the likely incorrect audio transcript; configuring the audio transcript correction machine learning model based on a training of a machine learning text-to-text transformer model using the transcript correction training data corpus; and executing the audio transcript correction machine learning model within a speech-to-text post-processing sequence of a speech-to-text service.