Automated Assistant Command Bundle Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users of automated assistants often utilize only a limited amount of the available functionality due to unawareness of performable actions and canonical commands, leading to inefficient use of network and computational resources as they receive irrelevant recommendations and must issue multiple commands to perform multiple actions.
Innovation Solution
The implementation of command bundles, which are groups of discrete actions that can be performed by an automated assistant, are generated and recommended to users based on historical data and user interactions, using machine learning models to determine relevance and optimize resource usage. These bundles can be activated by invocation phrases, contextual conditions, or graphical user interface elements, and may include slots with variable values that are resolved through user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If general recommendations for canonical commands are provided to users, then users may discover new actions, but network and computational resources are wasted on irrelevant recommendations
Solution Approach 1:
The system analyzes historical user interaction data to understand which actions users actually perform and uses this feedback to generate personalized command bundle recommendations. This ensures recommendations are relevant to each user's actual usage patterns, avoiding waste of network and computational resources on irrelevant suggestions while still expanding user awareness of available actions.
Solution Approach 2:
The system dynamically changes the parameters of recommendations based on user-specific historical data. Instead of providing static general recommendations, it adapts the content, timing, and type of command bundles suggested to each user based on their interaction history, device context, and usage patterns, thereby improving relevance and reducing resource waste.
2Productivity
If multiple disparate canonical commands are provided to perform multiple actions, then complete tasks can be achieved, but significant network and computational resources are consumed across multiple dialog turns
Solution Approach 1:
The system merges multiple related canonical commands into a single integrated command bundle that performs multiple actions atomically. By combining disparate commands that work together toward a common goal into one unified interaction, it enables complete task completion while significantly reducing network traffic and computational overhead compared to handling multiple separate dialog turns.
Solution Approach 2:
The system performs preliminary analysis of user intent and pre-assembles bundles of complementary actions before user invocation. By preparing these multi-action bundles in advance based on historical data and contextual understanding, the system can execute complex task sequences in a single user interaction, avoiding the need for multiple sequential dialog turns and reducing overall resource consumption.
3Adaptability or versatility
If command bundles are personalized based on historical data, then user engagement increases, but data processing and model computation requirements increase
Solution Approach 1:
The system segments the personalization process into distinct stages: historical data collection, pattern analysis, command bundle generation, and recommendation delivery. By dividing the complex data processing task into manageable segments that can be performed incrementally and cached, it achieves high personalization accuracy while controlling computational requirements at each stage.
Solution Approach 2:
The system creates simplified representations (embeddings) of user interaction patterns and command bundle characteristics that can be efficiently stored and compared. By copying essential features into compact vector representations rather than processing full interaction histories each time, it maintains high personalization accuracy while significantly reducing the computational complexity of real-time recommendation generation.
Data Source
AI summary
Generating and/or recommending command bundles for a user of an automated assistant. A command bundle comprises a plurality of discrete actions that can be performed by an automated assistant. One or more of the actions of a command bundle can cause transmission of a corresponding command and/or other data to one or more devices and/or agents that are distinct from devices and/or agents to which data is transmitted based on other action(s) of the bundle. Implementations determine command bundles that are likely relevant to a user, and present those command bundles as suggestions to the user. In some of those implementations, a machine learning model is utilized to generate a user action embedding for the user, and a command bundle embedding for each of a plurality of command bundles. Command bundle(s) can be selected for suggestion based on comparison of the user action embedding and the command bundle embeddings.