Voice Name Matching With Context Feedback for Call Initiation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Initiating communications between communication devices based on audio input can be challenging due to the potential for selecting incorrect names, leading to wasted bandwidth and processing resources.
Innovation Solution
A system and method that converts audio input into phonetic data, using machine learning algorithms to select a match from a database of names, incorporating non-phonetic data such as images, audio signatures, location, and role data to improve accuracy, and utilizes feedback for training to enhance future selections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If audio input is converted to phonetic data and matched against a database of names, then communication initiation becomes automated and convenient, but incorrect matches occur leading to wasted bandwidth and processing resources
Solution Approach 1:
The system collects feedback from users when a matched name is correct or incorrect, and uses this feedback to retrain the machine learning model. This continuous learning process improves match accuracy over time, reducing incorrect matches and the associated waste of bandwidth and processing resources while maintaining automated communication initiation.
2Measurement precision
If only phonetic data is used for name matching, then the system remains simple and fast, but accuracy decreases leading to incorrect matches
Solution Approach 1:
The system combines multiple types of data (phonetic data, non-phonetic audio characteristics, location data, role data) into a composite input for the machine learning model. This composite approach improves name matching accuracy by considering multiple factors simultaneously, while the model learns to weigh each factor appropriately, managing the complexity through intelligent integration rather than separate processing systems.
3Adaptability or versatility
If the system uses a large database of names with multiple potential matches, then coverage increases and more communications can be initiated, but the difficulty of selecting the correct name increases
Solution Approach 1:
The system moves beyond traditional phonetic matching by adding multiple dimensions of data including non-phonetic audio characteristics, location information, and role data. This multi-dimensional approach allows the machine learning model to distinguish between names that may sound similar by considering additional contextual factors, making correct name selection easier even when coverage over many potential matches is required.
Data Source
AI summary
A device, system and method to initiate a communication between communication devices based on audio input is provided. A computing device converts audio input from a first communication device into phonetic data corresponding to a target name from the audio input, and selects, using a machine learning algorithm, from names that potentially match the target name, a particular name as a match, using as input to the algorithm: the phonetic data; and non-phonetic data associated with one or more of the audio input, and at least one of the names. The computing device: initiates a communication between the first communication device, and a second communication device associated with the particular name; receives, from one or more of the first and second communication devices, feedback indicating whether the match of the particular name and the target name is positive or negative; and trains the machine learning algorithms using the feedback.