Speaker Diarization Correction via Text-Based Change Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speaker diarization systems often incorrectly detect speaker changes in audio streams, leading to errors where a change is falsely identified when none has occurred.

Innovation Solution

A method and system that corrects speaker diarization errors by performing speech-based speaker diarization, recognizing speech as text, detecting speaker changes based on the text using trained models like BERT, and subsequently correcting the diarization based on these detected changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If speaker diarization is performed using only speech-based methods, then the process is simple and fast, but speaker change detection accuracy deteriorates due to false detections

Engineering Contradiction:

Improvespeaker change detection accuracyVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges speech-based speaker diarization with text-based speaker change detection into a unified system. The speech diarization results are corrected by comparing against text analysis results, combining the speed of speech processing with the accuracy of text analysis to reduce false detections while maintaining system efficiency

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces text-based speaker change detection as an intermediary layer between speech input and final diarization output. This intermediary text analysis acts as a verification mechanism that corrects false speaker change detections from the speech-based system without requiring complete system redesign

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If text-based speaker change detection is added to correct speech-based diarization, then detection accuracy improves, but processing time increases

Engineering Contradiction:

Improvespeaker change detection precisionVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs speech-based speaker diarization first to identify potential speaker changes, then uses text-based detection only to verify and correct these pre-identified changes. This preliminary action approach avoids processing the entire audio stream through both methods, significantly reducing processing time while maintaining accuracy improvements

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies text-based speaker change detection partially - only to segments where speech-based diarization detected a speaker change. This partial application avoids the time cost of analyzing the entire transcript, achieving accuracy improvement with minimal additional processing time

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12315517B2Method and system for correcting speaker diarization using speaker change detection based on text

Publication Date: 2025.05.27 LINE WORKS CORP

AI summary

A method and system for correcting speaker diarization using a text-based speaker change detection. A speaker diarization correction method may include performing speaker diarization on an input audio stream; recognizing speech included in the input audio stream and converting the speech to text; detecting a speaker change based on the converted text; and correcting the speaker diarization based on the detected speaker change.