Hybrid Voice Messaging System with Human-in-the-Loop Transcription
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice messaging systems struggle to efficiently convert unstructured voicemail messages into text at a mass scale, particularly for large user bases, due to high costs and inefficiencies in processing times, and fail to accurately capture the meaning and idiomatic elements of messages, which is crucial for user confidence and practical application.
Innovation Solution
A hybrid system combining automatic speech recognition (ASR) with human operators and quality control, utilizing pre-processing, conversion resources, and quality control sub-systems to optimize human operator efficiency, and employing contextual information and language models to improve conversion accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional ASR systems are used to convert voicemail to text, then transcription can be automated, but the system fails to capture meaning and idiomatic elements accurately
Solution Approach 1:
The patent introduces human operators as an intermediary between the automated ASR system and the final text output. The ASR system provides initial transcription, which is then reviewed and refined by human operators who understand context, idioms, and meaning, thus combining automation with semantic accuracy
Solution Approach 2:
The system creates a composite transcription process combining machine-generated text from ASR with human editorial input. This composite approach leverages the speed and scalability of automation while incorporating human understanding of nuance, resulting in text that captures both efficiency and meaning
2Measurement precision
If human operators transcribe all messages, then accuracy improves, but cost and processing time increase prohibitively for mass scale
Solution Approach 1:
The system applies human review selectively rather than universally. Human operators review and refine only the portions of transcriptions that require contextual understanding or contain uncertain ASR output, while accepting straightforward transcriptions as-is, thus optimizing the balance between accuracy and processing capacity
Solution Approach 2:
The transcription process is segmented into multiple stages: initial automated ASR transcription, quality assessment of the ASR output, selective human review of problematic segments, and final assembly. This segmentation allows the system to handle mass volume while maintaining accuracy where needed
3Loss of time
If the system processes messages quickly to meet 2-5 minute turnaround, then user confidence improves, but processing complexity increases for large user bases
Solution Approach 1:
The system performs preliminary automated ASR transcription immediately upon receiving a voicemail, creating a draft text version before human review. This preliminary action ensures that even if human review takes time, a usable transcription is already available, meeting the fast turnaround requirement while preparing the message for subsequent quality enhancement
Solution Approach 2:
The system maintains continuous processing through parallel operations: ASR transcription occurs continuously for all incoming messages, while human operators continuously review and refine transcriptions. This continuous multi-stage processing ensures consistent fast turnaround times even as message volume scales to large user bases
Data Source
AI summary
A mass-scale, user-independent, device-independent, voice messaging system that converts unstructured voice messages into text for display on a screen is disclosed. The system comprises (i) computer implemented sub-systems and also (ii) a network connection to human operators providing transcription and quality control; the system being adapted to optimize the effectiveness of the human operators by further comprising 3 core sub-systems, namely (i) a pre-processing front end that determines an appropriate conversion strategy; (ii) one or more conversion resources; and (iii) a quality control sub-system.