Single-Request Multi-Person Calls with Confidence-Based Recipient Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in disambiguating ambiguous user requests for multi-person calls, particularly in determining the intended recipients without additional user input.
Innovation Solution
A personalized machine-learning model is used to determine candidate users for a multi-person call, associating each with a confidence score, and confirms the selection with user input when scores are below a threshold, leveraging user profile information, historical interactions, and contextual data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single request is used for multi-person calling, then ease of operation is improved, but reliability deteriorates due to ambiguous recipient identification
Solution Approach 1:
The system implements feedback by providing confidence scores for each candidate recipient and seeking user confirmation when the top candidate's confidence score falls below a threshold. This feedback loop allows the system to verify ambiguous requests before execution, resolving the contradiction between ease of operation and reliability by maintaining simple single-request input while adding a confirmation step only when necessary.
Solution Approach 2:
The patent introduces an intermediary confirmation mechanism that acts as a mediator between the simple single-request input and the reliable multi-person call execution. When ambiguity is detected (low confidence score), the intermediary confirmation step bridges the gap by presenting candidate recipients to the user for verification, thus maintaining both ease of operation and reliability.
2Measurement precision
If candidate users are determined using a personalized machine-learning model, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system applies partial action by using the machine-learning model only when necessary - specifically, when the top candidate's confidence score is below the threshold. In cases of high confidence, the system proceeds directly without invoking the complex model for full candidate generation and sorting, thus achieving high measurement precision when needed while minimizing the impact on device complexity.
Solution Approach 2:
The patent changes the parameter of confidence score threshold to control when the machine-learning model is activated. By adjusting this threshold parameter, the system can balance between measurement precision (higher threshold activates model more often) and device complexity (lower threshold reduces model activation frequency), allowing optimization based on specific operational requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In one embodiment, a method includes receiving a user request to call a plurality of second users simultaneously from a client system associated with a first user via an assistant xbot, wherein the user request comprises one or more references referring the plurality of second users, identifying the plurality of second users by resolving the one or more references to a plurality of entity identifiers associated with the plurality of second users, respectively, and initiating a group call to the plurality of second users simultaneously based on the plurality of entity identifiers by the assistant xbot.