Automated Assistant Dialog Context Transition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated assistants lose dialog context when users switch topics or interact with third-party applications, requiring users to restart conversations and re-input information, which is inefficient and resource-intensive.
Innovation Solution
The system preserves multiple semantically distinct dialog contexts and allows users to transition between them seamlessly by using transition commands, such as 'Hey assistant, let's go back to,' or selecting a 'back button,' enabling users to resume previous conversations without losing context.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the automated assistant preserves only the most recent dialog context, then memory usage is minimized, but the user cannot resume previous conversations after switching topics or applications
Solution Approach 1:
The dialog history is segmented into multiple distinct contexts or topics, each stored separately. When the user switches topics or applications, the system identifies and preserves the relevant context segment, allowing the user to return to it later without loading the entire dialog history into memory.
Solution Approach 2:
The system performs preliminary actions by storing and tagging dialog contexts in advance with metadata indicating their topic or semantic meaning. This allows quick retrieval and resumption of specific contexts without requiring the system to maintain all contexts in active memory simultaneously.
2Loss of information
If the automated assistant maintains multiple dialog contexts in memory, then users can resume previous conversations, but computational resources and energy consumption increase
Solution Approach 1:
The system extracts only the essential elements of dialog contexts (intents, slot values, key entities) and stores them in a compact format. Full dialog transcripts are not maintained in memory, but rather summarized representations that can be quickly reconstructed when needed, reducing both memory usage and energy consumption.
Solution Approach 2:
The system changes the representation parameters of dialog contexts from full transcripts to structured summaries containing only critical information (intents, slots, entities). This parameter transformation reduces the computational burden while preserving the ability to resume conversations effectively.
3Reliability
If the automated assistant requires users to re-input information when resuming conversations, then processing accuracy is maintained, but user convenience and productivity decrease
Solution Approach 1:
The system provides feedback to the user by presenting the stored dialog context summary when a user attempts to resume a conversation. The user can then confirm or correct the pre-filled information, combining automated context retrieval with user verification to maintain accuracy while improving convenience.
Solution Approach 2:
The system performs preliminary action by pre-processing and storing dialog contexts with structured information (intents, slots, entities) that can be automatically reinstated. This eliminates the need for users to re-input information, while still allowing for verification and correction to maintain processing accuracy.
4Device complexity
If the automated assistant uses simple context persistence, then implementation complexity is low, but the system cannot handle users returning from third-party applications after significant time delays
Solution Approach 1:
The system performs preliminary action by continuously tagging and metadata-ing dialog contexts with topic identifiers and temporal information. When users return after delays, the system can quickly search and retrieve relevant contexts based on these pre-established tags, rather than requiring complex real-time analysis of the entire dialog history.
Solution Approach 2:
The system introduces an intermediary layer of context management that bridges simple persistence and complex real-time analysis. Dialog contexts are stored with structured metadata that acts as an intermediary index, enabling efficient retrieval without requiring the system to maintain complex active memory structures or perform computationally intensive searches.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Techniques are described related to prior context retrieval with an automated assistant. In various implementations, instance(s) of free-form natural language input received from a user during a human-to-computer dialog session between the user and an automated assistant may be used to generate a first dialog context. The first dialog context may include intent(s) and slot value(s) associated with the intent(s). Similar operations may be performed with additional inputs to generate a second dialog context that is semantically distinct from the first dialog context. When a command is received from the user to transition the automated assistant back to the first dialog context, natural language output may be generated that conveys at least one or more of the intents of the first dialog context and one or more of the slot values of the first dialog context. This natural language output may be presented to the user.