Audio-Derived Text Error Correction for Poor-Quality Speech

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing automatic speech recognition (ASR) systems face significant interference due to poor audio quality, leading to errors in speech recognition results, which negatively impact user experience and interaction efficiency.

Innovation Solution

A text processing method and apparatus that dynamically corrects errors in text generated from audio information by identifying to-be-error-corrected words and determining corresponding replacement words, using various error correction types such as text-pronunciation-based and text-content-based methods, and updating the text accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If automatic speech recognition is applied to recognize speech from audio stream, then speech recognition capability is provided, but recognition accuracy deteriorates when audio quality is poor

Engineering Contradiction:

Improvespeech recognition accuracyVSAvoidinterference from poor audio quality

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces text information as an intermediary between audio information and ASR results. The text information serves as a mediator that can be processed through error correction mechanisms (identifying to-be-error-corrected words and determining replacement words) to improve the final recognition accuracy, thereby resolving the contradiction between providing speech recognition capability and maintaining accuracy under poor audio conditions

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements a feedback mechanism where ASR results are converted to text information, which then undergoes error correction processing. The corrected text information feeds back into the system to improve subsequent recognition accuracy, creating a closed-loop system that continuously improves performance despite poor audio quality inputs

Inventive Principle:

Principle #23Feedback

2Measurement precision

If error correction is performed on text information, then text accuracy is improved, but processing time increases

Engineering Contradiction:

Improvetext accuracyVSAvoiderror correction processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary error correction processing on text information before final ASR output generation. By identifying to-be-error-corrected words and determining replacement words in advance, the system prepares corrected text information that can be quickly output, reducing the perceived processing time while maintaining high accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies error correction selectively to text information that requires it, rather than processing all text uniformly. The system identifies specific to-be-error-corrected words and applies correction only where needed, optimizing the balance between improvement and processing time by avoiding unnecessary full-text processing

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12406674B2Text processing method and apparatus, electronic device, and medium

Publication Date: 2025.09.02 DOUYIN VISION CO LTD

AI summary

Provided are a text processing method and apparatus, an electronic device, and a medium. The method includes the following: target text information generated based on audio information is acquired; a to-be-error-corrected word in the target text information and a target candidate replacement word corresponding to the to-be-error-corrected word are determined; and a target replacement word corresponding to the to-be-error-corrected word is determined according to the target candidate replacement word, and the target text information is updated based on the target replacement word.