Word-level speech correction via word lattice selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Transcription of speech input into text on mobile computing devices often results in errors due to limited user interfaces, making it difficult for users to correct mistakes efficiently.
Innovation Solution
A system that sends speech audio data to a transcription system, which generates a word lattice including probable transcriptions and alternate words, allowing users to select and replace incorrect words through a limited interface, such as a touchscreen, with minimal user inputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If speech input is transcribed on mobile devices with limited interfaces, then transcription speed is improved, but correction accuracy deteriorates
Solution Approach 1:
The system performs preliminary actions by generating a word lattice with multiple alternate transcriptions before the user needs to correct errors. The transcription system creates multiple possible word sequences and presents them in advance, allowing the user to select from pre-computed alternatives rather than typing corrections from scratch.
Solution Approach 2:
The word lattice serves as an intermediary structure between the speech recognition system and the user interface. It contains multiple alternate transcriptions with probability weights, acting as a mediator that provides correction options without requiring complex user input or additional communication with the transcription system.
2Manufacturing precision
If users correct transcribed text on mobile devices, then text accuracy is improved, but user effort increases
Solution Approach 1:
The system applies partial action by only requiring the user to select from pre-generated alternate words rather than performing full correction operations. The word lattice already contains multiple alternate transcriptions, so the user only needs to choose the correct one rather than typing or navigating complex correction menus.
Solution Approach 2:
The transcription system effectively serves itself by generating multiple alternate transcriptions and probability weights without user intervention. The system autonomously creates the word lattice with alternative word sequences, eliminating the need for users to manually request or navigate through multiple correction iterations.
3Manufacturing precision
If multiple correction iterations are performed, then text accuracy is improved, but communication overhead increases
Solution Approach 1:
The system performs preliminary actions by generating the complete word lattice with multiple alternate transcriptions and probability weights in a single initial communication round. This eliminates the need for multiple back-and-forth communication iterations between the mobile device and transcription system, as all correction options are available locally from the first response.
Data Source
AI summary
The subject matter of this specification can be implemented in, among other things, a computer-implemented method for correcting words in transcribed text including receiving speech audio data from a microphone. The method further includes sending the speech audio data to a transcription system. The method further includes receiving a word lattice transcribed from the speech audio data by the transcription system. The method further includes presenting one or more transcribed words from the word lattice. The method further includes receiving a user selection of at least one of the presented transcribed words. The method further includes presenting one or more alternate words from the word lattice for the selected transcribed word. The method further includes receiving a user selection of at least one of the alternate words. The method further includes replacing the selected transcribed word in the presented transcribed words with the selected alternate word.