Voice Correction Suggestions via Phonetic Biasing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice recognition systems often misrecognize user inputs, leading to incorrect search queries and results, requiring users to restart the voice recognition process or manually correct errors, which is inefficient and time-consuming.
Innovation Solution
A method that allows users to select and correct misrecognized terms in voice recognition outputs by inputting additional letters, with suggested corrections ranked based on phonetic similarity and popularity, enabling quick identification of intended queries without reinitiating the voice recognition process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional speech to text processes are used for voice recognition, then the device can respond to speech inputs, but misrecognitions occur leading to incorrect results
Solution Approach 1:
The system displays the recognition output to the user and receives feedback through user inputs (letters or corrections). This feedback loop allows the system to adjust and refine the recognized output, transforming misrecognitions into correct interpretations by incorporating user corrections into the final result.
Solution Approach 2:
The system introduces an intermediary correction interface between the initial voice recognition output and the final search query. This intermediary layer allows users to selectively correct misrecognized terms without completely rejecting the voice input, bridging the gap between automated recognition and accurate user intent.
2Reliability
If users manually correct misrecognized terms by reinitiating voice recognition, then accuracy can be improved, but time is lost
Solution Approach 1:
The system performs preliminary corrections by displaying the recognition output and allowing users to make minor adjustments (inputting a few letters) before finalizing the query. This preliminary action prevents the need to completely reinitiate the voice recognition process, saving time while maintaining accuracy.
Solution Approach 2:
Instead of requiring complete re-recognition of the entire query, the system allows users to make partial corrections by inputting only the necessary letters or correcting specific misrecognized terms. This partial action approach corrects errors efficiently without the excessive time cost of full re-recognition.
3Adaptability or versatility
If suggested corrections are provided without bias, then all possibilities are shown, but user intent is not quickly identified
Solution Approach 1:
The system applies different weighting or bias to different correction suggestions based on local characteristics such as phonetic similarity, common usage patterns, or context. This local quality adjustment prioritizes the most likely intended query in the suggestion list, allowing users to quickly identify their intent without examining all possible corrections equally.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for natural language processing. One of the method includes receiving a voice input from a user device; generating a recognition output; receiving a user selection of one or more terms in the recognition output; receiving a user input of one or more letters replacing the user selected one or more terms; determining suggested correction candidates based in part on the user input and the voice input; and providing one or more suggested correction candidates to the user device as suggested corrected recognition outputs.