Speech Segmentation Using Pause Detection and Diarization
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-laden text inputs that result in misleading insights during analysis.
Innovation Solution
An apparatus and method that utilize a combination of pause detection techniques and speaker diarization to segment speech audio into more accurate units, employing multiple threads to identify sentence pauses and speaker changes, and then use acoustic and language models to generate transcripts, with dynamic weighting of models based on noise levels and uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional speech-to-text conversion is used without segmentation, then processing is simpler and faster, but accuracy deteriorates due to environmental noise, dialects, accents, and pronunciation variations
Solution Approach 1:
The patent applies segmentation by dividing speech audio into smaller units using pause detection to identify sentence boundaries and speaker diarization to separate different speakers. This segmentation allows each segment to be processed independently with acoustic and language models, improving accuracy by reducing the complexity of processing entire speech streams at once while handling noise, dialects, and accents more effectively in smaller contexts.
2Measurement precision
If multiple pause detection techniques and speaker diarization are employed to segment speech audio, then speech-to-text conversion accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by performing pause detection and speaker diarization before the main speech-to-text conversion process. These preprocessing steps segment the audio into manageable units in advance, allowing the acoustic and language models to process pre-segmented data more efficiently. This upfront segmentation reduces the computational burden during the main transcription phase, balancing accuracy improvement with reasonable processing time.
3Productivity
If speech audio is segmented without accurate pause detection, then processing is faster and simpler, but sentence completeness deteriorates leading to errors in transcription
Solution Approach 1:
The patent applies feedback by using multiple pause detection techniques that analyze acoustic features and provide feedback about potential sentence boundaries. The system integrates results from different detection methods, using confidence scores and contextual information to confirm or reject pause locations. This feedback mechanism ensures that segments contain complete sentences while maintaining processing efficiency through automated decision-making based on accumulated evidence from multiple detection approaches.
Data Source
AI summary
A system, method, and computer-program product includes constructing a transcript adaptation training data corpus that includes a plurality of transcript normalization training data samples, wherein each of the plurality of transcript normalization training data samples includes: a predicted audio transcript that includes at least one numerical expression, an adapted audio transcript that includes an alphabetic representation of the at least one numerical expression, and a transcript normalization identifier that, when applied to a model input comprising a target audio transcript, defines a text-to-text transformation objective causing a numeric-to-alphabetic expression machine learning model to predict an alphabetic-equivalent audio transcript that represents each numerical expression included in the target audio transcript in one or more alphabetic tokens; configuring the numeric-to-alphabetic expression machine learning model based on a training of a machine learning text-to-text transformer model using the transcript adaptation training data corpus; and executing the numeric-to-alphabetic expression machine learning model.