Automated Assistant Invocation Routing for Accurate Activation

Solution Overview

Problem

Existing automated assistants often require explicit invocation and are prone to incorrect activation, leading to unnecessary resource consumption and increased processing time due to multiple models monitoring for invocation phrases, which can result in false positives and inefficient resource utilization.

Innovation Solution

A general automated assistant processes invocation inputs to determine which of multiple candidate assistants to invoke based on features such as prosodic characteristics, location, and additional context, reducing the need for multiple models and minimizing incorrect activations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multiple automated assistants monitor for invocation phrases simultaneously, then each assistant can be independently activated, but false positives increase and resource consumption increases

Engineering Contradiction:

Improveactivation accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

Multiple automated assistants share a single invocation model instead of each maintaining separate models. The invocation model is trained to recognize invocation phrases for multiple assistants, and a routing mechanism determines which assistant should handle the invocation based on the detected phrase and contextual factors.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

A single invocation model serves multiple automated assistants universally, detecting invocation phrases that can trigger different assistants. The system becomes multi-functional by handling invocations for various assistants through one shared model rather than requiring dedicated models for each assistant.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If multiple automated assistants monitor for invocation phrases, then comprehensive coverage is achieved, but processing time increases due to multiple models evaluating each audio input

Engineering Contradiction:

Improveassistant selection capabilityVSAvoidinvocation processing time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

Multiple automated assistants share a single invocation model instead of each maintaining separate models. The invocation model is trained to recognize invocation phrases for multiple assistants, and a routing mechanism determines which assistant should handle the invocation based on the detected phrase and contextual factors.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system pre-trains a single invocation model to recognize multiple invocation phrases associated with different assistants before runtime. This preliminary training enables the model to efficiently route invocations to the appropriate assistant without requiring multiple models to evaluate each input during operation.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If each automated assistant has its own invocation model, then activation precision is maintained, but device complexity increases

Engineering Contradiction:

Improveinvocation detection accuracyVSAvoidnumber of invocation models

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

Multiple automated assistants share a single invocation model instead of each maintaining separate models. The invocation model is trained to recognize invocation phrases for multiple assistants, and a routing mechanism determines which assistant should handle the invocation based on the detected phrase and contextual factors.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

A single invocation model serves multiple automated assistants universally, detecting invocation phrases that can trigger different assistants. The system becomes multi-functional by handling invocations for various assistants through one shared model rather than requiring dedicated models for each assistant.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250349297A1Selecting between multiple automated assistants based on invocation properties

Publication Date: 2025.11.13 GOOGLE LLC

AI summary

Systems and methods for determining, based on invocation input that is common to multiple automated assistants, which automated assistant to invoke in lieu of invoking other automated assistants. The invocation input is processed to determine one or more invocation features that may be utilized to determine which, of a plurality of candidate automated assistants, to invoke. Further, additional features are processed that can indicate which, of the plurality of invocable automated assistants, to invoke. Once an automated assistant has been invoked, additional audio data and/or features of additional audio data are provided to the invoked automated assistant for further processing.