Personalized Speech Recognition Overrides for Noisy Transcription
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic speech recognition (ASR) engines struggle with accurately transcribing user utterances due to background noise and lack of personalized override mechanisms, often providing binary and non-tailored corrections.
Innovation Solution
An acoustic model trained with historical user utterance data and override triggering rules to improve transcription accuracy by identifying and correcting common errors using machine learning-based classification models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a binary override list is used to correct transcribed words, then common errors can be corrected, but the system lacks personalization and cannot adapt to individual user speech patterns
Solution Approach 1:
The system performs preliminary actions by collecting historical user utterance data and training a personalized acoustic model before actual speech recognition occurs. The acoustic model is pre-trained with user-specific data to enable personalized correction capabilities without adding complexity during real-time operation.
Solution Approach 2:
The system implements feedback mechanisms by using historical user utterance data to continuously improve the acoustic model. The model learns from past corrections and user speech patterns, adapting to individual users over time while maintaining a relatively simple operational structure.
2Measurement precision
If an acoustic model trained with historical data is used to override erroneous transcriptions, then transcription accuracy improves, but the system complexity increases
Solution Approach 1:
The system segments the speech recognition process into distinct components: a basic ASR engine for initial transcription and a specialized acoustic model for correction. This segmentation allows the complex acoustic model to operate independently on specific problematic cases without complicating the entire system architecture.
Solution Approach 2:
The acoustic model serves as an intermediary between the basic ASR engine and the final transcription output. It acts as a specialized mediator that intervenes only when correction is needed, improving accuracy without requiring the entire system to be complex.
3Reliability
If general ASR engines are used for all speech transcription, then the system remains simple and broadly applicable, but transcription accuracy deteriorates due to inability to handle common error cases
Solution Approach 1:
The system applies local quality by making the acoustic model user-specific rather than universal. Each user receives personalized correction capabilities tailored to their speech patterns and common errors, improving reliability for individual users without requiring a completely different system for each user.
Data Source
AI summary
Methods, systems, and apparatuses for improved speech recognition and transcription of user utterances are described herein. A user utterance may be processed by a speech recognition computing device. One or more acoustic features associated with the user utterance may be used to determine whether one or more actions are to be performed based on a transcription of the user utterance.