Automated Assistant Predicted-Action Pre-Caching to Reduce Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated assistants exhibit latency due to extensive computational and network processing of spoken utterances, leading to delayed responses in performing actions.
Innovation Solution
Predicting and initializing subtasks of user actions before they are explicitly requested, using machine learning models to generate action predictions and perform necessary subtasks proactively, thereby reducing latency and resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If spoken utterances are processed through speech-to-text and natural language processing to perform automated assistant actions, then the assistant can understand and execute user commands, but latency increases due to extensive computational and network processing
Solution Approach 1:
The system performs preliminary actions by predicting user intents and pre-processing data before explicit user requests. Action prediction models generate predicted actions based on current context, and subtasks are initialized in advance (e.g., pre-fetching data, establishing connections) so that when the user actually requests the action, the processing is already complete or near-complete, significantly reducing perceived latency while maintaining accurate command execution
2Measurement precision
If extensive computational processing is performed to accurately interpret spoken utterances and determine commands, then command accuracy improves, but computational resources are consumed and processing time increases
Solution Approach 1:
The system performs preliminary action by using action prediction models to anticipate user intents before explicit requests. These models analyze current context (screen state, application data, user behavior patterns) to generate predicted actions, allowing the system to pre-process and prepare responses in advance. This reduces the need for extensive real-time computational processing when actual user commands are given, thereby maintaining high accuracy while reducing instantaneous resource consumption
Solution Approach 2:
The system applies partial action by selectively processing only the most likely predicted actions rather than exhaustively analyzing all possible interpretations. The action prediction model ranks potential actions by probability, and the system focuses computational resources on the top predictions, performing extensive processing only when necessary to maintain accuracy for high-probability intents while using lighter processing for lower-probability cases
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Implementations herein relate to pre-caching data, corresponding to predicted interactions between a user and an automated assistant, using data characterizing previous interactions between the user and the automated assistant. An interaction can be predicted based on details of a current interaction between the user and an automated assistant. One or more predicted interactions can be initialized, and/or any corresponding data pre-cached, prior to the user commanding the automated assistant in furtherance of the predicted interaction. Interaction predictions can be generated using a user-parameterized machine learning model, which can be used when processing input(s) that characterize a recent user interaction with the automated assistant. Should the user command the automated assistant in a way that is aligned with a pre-cached, predicted interaction, the automated assistant will exhibit instant fulfillment of the command, thereby eliminating any latency that the user would have otherwise experienced interacting with the automated assistant.