Audio Personal Information Masking via Transcription and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods are ineffective in protecting personal information in audio recordings, particularly due to the dynamic nature of personal data and the difficulty in manually identifying and masking sensitive information within these recordings.
Innovation Solution
The system employs machine learning classifiers and Regular Expressions to identify and mask personal information in audio recordings by transcribing the audio into text, using learned patterns to classify and replace sensitive information with white noise, ensuring it remains unrecoverable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual methods are used to identify and mask personal information in audio recordings, then the masking process can be performed, but the effectiveness is insufficient due to the dynamic nature of personal data and the difficulty in manually identifying sensitive information
Solution Approach 1:
The patent replaces manual identification and masking operations with an automated system comprising audio transcription, text analysis using regular expressions, machine learning classification, and automated audio masking. This substitution of mechanical/manual processes with automated computational systems directly resolves the contradiction by improving reliability through consistent automated application while eliminating the operational difficulty of manual identification.
Solution Approach 2:
The system performs self-service by automatically transcribing audio, identifying personal information patterns, classifying sensitive content, and applying masking without human intervention. The automated pipeline serves itself by processing audio recordings end-to-end, improving both reliability through consistent execution and ease of operation by eliminating manual effort.
2Productivity
If automated machine learning classifiers and Regular Expressions are used to identify and mask personal information, then the productivity and coverage of personal information detection is improved, but the device complexity increases
Solution Approach 1:
The patent segments the personal information detection and masking system into distinct functional modules: audio transcription module, regular expression pattern matching module, machine learning classification module, and audio masking module. This segmentation allows each component to be independently optimized and managed, improving overall productivity while making the complex system more tractable and maintainable.
Solution Approach 2:
The system employs universal components that handle multiple aspects of personal information protection. The machine learning classifier serves both identification and classification functions, while the regular expression engine handles pattern matching across different types of personal information. This multi-functionality improves productivity by reducing the number of separate systems needed while managing complexity through standardized universal components.
Data Source
AI summary
One example method includes transcribing a portion of the audio component to create a transcription file that includes text, searching the text of the transcription file and identifying information in the text that may include personal information, defining a textual window that includes the information, evaluating the text in the textual window to identify personal information, and masking the personal information in the audio component of the recording. The personal information may be masked with information of a non-personal nature.