Speech-to-Text Profile Training for Relay Transcription
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current telephone relay systems require communication assistants to transcribe speech into text, which is inefficient and reduces privacy for hard of hearing individuals, as they need assistance for every call, and may introduce errors in transcription.
Innovation Solution
A speech-to-text profile is created and trained for each call participant, allowing for automatic transcription of audio voice signals into text without the need for a communication assistant, using a unique identifier and continuous training during calls to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If communication assistants manually transcribe speech into text, then transcription accuracy can be maintained, but the process is inefficient and reduces privacy for hard of hearing individuals
Solution Approach 1:
The system enables self-service by automatically transcribing speech into text without requiring human communication assistants. The speech-to-text system processes audio signals autonomously, converting spoken words to text automatically, which eliminates the need for manual transcription while maintaining efficiency and privacy for hard of hearing individuals
Solution Approach 2:
The patent replaces the mechanical manual transcription process with an automated electronic speech-to-text system. The system uses audio signal processing and text generation algorithms to automatically convert speech into text, substituting the human operator's mechanical typing process with an automated digital system that operates continuously without human intervention
2Reliability
If communication assistants are required for every call, then transcription can be performed, but it introduces errors and reduces privacy
Solution Approach 1:
The system eliminates the need for human communication assistants by implementing automated speech-to-text functionality. The system processes audio signals and generates text automatically, reducing system complexity by removing the human intermediary while improving reliability through consistent automated processing without human error
Solution Approach 2:
The patent introduces an automated speech-to-text system as an intermediary between the audio signal and text output. This intermediary process automatically converts speech to text without requiring human intervention, thereby improving reliability by eliminating human error while managing system complexity through automated processing
3Loss of time
If manual transcription is used, then transcription can be performed, but it is time-consuming and less private
Solution Approach 1:
The system performs self-service by automatically transcribing speech into text without requiring human operators. The automated speech-to-text system processes audio signals continuously and converts them to text in real-time, significantly reducing transcription time while maintaining operational simplicity through automatic processing
Solution Approach 2:
The patent implements continuous transcription by processing audio signals in real-time as they are captured. The speech-to-text system operates continuously during the call, converting speech to text without interruption or delay, thereby eliminating the time loss associated with manual transcription while maintaining ease of operation through automated continuous processing
Data Source
AI summary
An illustrative method includes receiving, at a processor of a computing device, an audio voice signal of a first call participant during a first call, where the first call is a communication across a communication network. The method further includes determining an identity of the first call participant and determining a speech to text profile associated with the identity of the first call participant, where the speech to text profile includes at least one rule for transcribing a word in the audio voice signal into text. The method further includes generating a text output, where the text output is a transcribed version of a plurality of words identified in the audio voice signal of the first call participant. At least one of the plurality of words identified is identified using the at least one rule.