Conversational Agent Multi-Task Orchestration via Shared State
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current conversational agents lack the ability to share slot values across tasks, express complex constraints on user goals, and interleave discussions of multiple related tasks, limiting their efficiency and natural interaction with users.
Innovation Solution
A method and system that estimate task sets based on user input utterances, assign confidence scores, and select the most relevant tasks to generate responses, enabling seamless orchestration of multiple domain-specific systems for multi-domain, multi-task interactions with complex goals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If current conversational agents use independent execution of domain-specific task scripts, then each task can be handled individually, but information sharing across tasks is minimal and users must provide common slot values separately for each task
Solution Approach 1:
The patent merges multiple independent domain-specific task scripts into a unified task execution framework. The dialog state tracker maintains a shared state across tasks, allowing slot values to be automatically propagated between related tasks. For example, when a user books a flight, the destination and dates are automatically shared with the hotel booking task, eliminating redundant user input and enabling seamless multi-task handling.
Solution Approach 2:
The patent implements a universal dialog state tracker that serves multiple domain-specific tasks simultaneously. This tracker maintains a consolidated representation of user goals, constraints, and slot values that can be accessed by any task. The system enables a single conversational agent to handle diverse tasks (flight booking, hotel reservation, restaurant reservation) using a common state management mechanism, improving adaptability without proportionally increasing complexity.
2Adaptability or versatility
If current conversational agents handle single tasks with simple user goals, then task execution is straightforward, but users cannot communicate goals related to multiple tasks in a single utterance or provide multiple preferential constraints
Solution Approach 1:
The patent implements a dynamic dialog state tracker that adapts its structure and content based on the complexity of user goals. The tracker can handle both simple single-task goals and complex multi-task constraints by dynamically adjusting the task set and associated slot values. When users provide preferential constraints (e.g., 'prefer hotels under $200'), the system dynamically updates the state to reflect these preferences across multiple tasks, allowing flexible expression of complex requirements without overwhelming the user.
Solution Approach 2:
The system provides feedback mechanisms that allow users to refine and adjust their goals during conversation. The dialog state tracker continuously updates based on user input, confirming understanding of complex constraints and allowing users to modify preferences. This feedback loop enables users to express complex multi-task goals iteratively, maintaining ease of operation while handling sophisticated requirements.
3Productivity
If current conversational agents execute tasks sequentially without interleaving, then task management is simple, but related tasks cannot be interleaved to manage shared constraints efficiently
Solution Approach 1:
The patent implements preliminary action by identifying and extracting all relevant slot values and constraints from user input before executing individual tasks. The dialog state tracker pre-processes the user's multi-task goal, determining the complete set of tasks and their interdependencies upfront. This allows the system to efficiently interleave task execution while managing shared constraints, as the overall task structure and resource requirements are already established before detailed task processing begins.
Solution Approach 2:
The patent adds a temporal dimension to task execution by allowing flexible interleaving of tasks based on shared constraints and user preferences. Instead of rigid sequential execution, the system can switch between tasks dynamically, executing sub-tasks in an optimal order that respects shared resources and constraints. For example, the system might book flights first, then use the travel dates to inform hotel reservations, demonstrating multi-dimensional task orchestration that improves productivity without excessive complexity.
Data Source
AI summary
The present teaching relates to facilitating a guided dialog with a user. In one example, an input utterance is obtained from the user. One or more task sets are estimated based on the input utterance. Each of the one or more task sets includes a plurality of tasks estimated to be requested by the user via the input utterance and is associated with a confidence score computed based on statistics with respect to the plurality of tasks in the task set. At least one of the one or more task sets is selected based on their respective confidence scores. A response is generated based on the tasks in the selected at least one task set. The response is provided to the user.