Context-Aware Disambiguation for Voice Command Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition technologies accurately identify voice inputs but struggle to execute intended actions due to ambiguities in commands, such as determining which file to send or which contact to send it to, as they lack the ability to leverage context data for disambiguation.
Innovation Solution
The proposed solution involves identifying ambiguous words in user voice inputs and using context data, both within and external to the input, to disambiguate commands, allowing the system to determine the appropriate action to be executed, such as accessing contact lists or file associations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional speech recognition technology is used to accurately identify voice inputs, then word recognition accuracy is improved, but command execution reliability deteriorates due to ambiguities in interpreting intended actions
Solution Approach 1:
The system performs preliminary disambiguation by accessing context data before executing the command. It identifies ambiguous words in the transcribed text and resolves them by comparing with context data (such as contact lists, file associations, or previously used commands) to determine the most likely intended meaning, ensuring reliable command execution before action is committed.
Solution Approach 2:
Context data serves as an intermediary between the recognized words and the intended command execution. The system uses context data (contact lists, file associations, application states) as a mediator to resolve ambiguities in words like 'send', 'open', or specific entity references, bridging the gap between literal word recognition and intended user action.
2Reliability
If context data is accessed to disambiguate commands, then command execution reliability is improved, but device complexity increases due to additional data access and processing requirements
Solution Approach 1:
The system uses self-service by leveraging context data that is already available within the information handling device. Instead of requiring external assistance or complex additional hardware, the processor utilizes existing context data (contact lists, file associations, application states) stored in the device's memory to resolve ambiguities, making the system self-sufficient in disambiguation.
Solution Approach 2:
Context data is prepared and made available in advance within the device's memory structures. The system performs preliminary organization of context data (contact lists, file associations) so that when disambiguation is needed, the processor can efficiently query and compare against pre-organized data, reducing the complexity of real-time processing.
3Ease of operation
If ambiguous words are resolved using context data, then user input requirements are reduced, but processing time increases due to additional context analysis
Solution Approach 1:
The system performs partial context analysis by focusing only on words that are actually ambiguous in the given context. Instead of analyzing the entire command structure or all possible interpretations, it identifies specific ambiguous words (such as verbs with multiple meanings or entity references) and performs targeted disambiguation only for those words, reducing unnecessary processing time while still achieving accurate interpretation.
Data Source
AI summary
An embodiment provides a method, including: receiving, at an audio receiver of an information handling device, user voice input; identifying, using a processor, words included in the user voice input; determining, using the processor, one of the identified words renders ambiguous a command included in the user voice input; accessing, using the processor, context data; disambiguating, using the processor, the command based on the context data; and committing, using the processor, a predetermined action according to the command. Other aspects are described and claimed.