ASR Pronunciation Prediction Using Correction Feedback

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current Automatic Speech Recognition (ASR) systems face challenges in accurately predicting pronunciations, especially for proper nouns, unfamiliar words, and varying speech contexts, leading to the need for continuous correction and updating of pronunciation models.

Innovation Solution

The method involves selecting a predicted text sample from an audio sample, receiving a correction text sample, updating the encoding of allowable pronunciations based on the predicted and correction text samples, and predicting the pronunciation of the correction text sample using the updated encoding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If ASR systems use traditional pronunciation models, then the system structure remains simple, but the accuracy of pronunciation prediction deteriorates for proper nouns and unfamiliar words

Engineering Contradiction:

Improvepronunciation prediction accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system implements feedback by using correction text samples to update the pronunciation model. When users correct predicted text, the system incorporates these corrections into the encoding of allowable pronunciations, creating a continuous improvement loop that enhances accuracy without requiring complete system redesign

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary actions by pre-computing and storing encodings of allowable pronunciations for correction text samples. This preprocessing enables faster real-time pronunciation prediction while maintaining the ability to adapt to new words and proper nouns

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If ASR systems maintain fixed pronunciation models, then the model structure remains stable, but the adaptability to varying speech contexts and new words deteriorates

Engineering Contradiction:

Improveadaptability to speech variabilityVSAvoidmodel structure stability

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The system applies dynamics by making the pronunciation model updateable and adaptable. The encoding of allowable pronunciations can be dynamically modified based on correction text samples, allowing the system to adapt to new words, proper nouns, and varying speech contexts while maintaining operational stability through structured update mechanisms

Inventive Principle:

Principle #15Dynamics

3Measurement precision

If ASR systems do not incorporate correction inputs, then the processing speed remains fast, but the pronunciation prediction accuracy deteriorates

Engineering Contradiction:

Improvepronunciation prediction accuracyVSAvoidmodel updating time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-computing encodings of allowable pronunciations during off-peak times or in the background. This allows correction incorporation without significantly impacting real-time processing speed, as the heavy computational work is done in advance

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250046296A1Automated prediction of pronunciation of text entities based on prior prediction and correction

Publication Date: 2025.02.06 GOOGLE LLC

AI summary

A method, device, and computer-readable storage medium for predicting pronunciation of a text sample. The method includes selecting a predicted text sample corresponding to an audio sample, receiving a correction text sample corresponding to the audio sample, updating an encoding of allowable pronunciations of the correction text sample based on the predicted text sample and the audio sample, the updated encoding of allowable pronunciations of the correction text sample including a pronunciation of the predicted text sample, and predicting a pronunciation of the correction text sample based on the updated encoding of allowable pronunciations of the correction text sample.