Multi-turn Canned Dialog Routing for Intent Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital assistants face challenges in efficiently handling user input due to the computational intensity and complexity of interpreting natural language, particularly in identifying domains and intents, which hinders quick deployment and robustness.
Innovation Solution
A method utilizing a computationally lightweight, predefined dialog scheme in conjunction with a sophisticated natural-language processing system to handle user input, allowing for quick deployment while maintaining a robust architecture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a sophisticated natural-language processing system is used to interpret user intent and identify domains, then the assistant can handle complex user requests accurately, but the computational demand and implementation complexity increase significantly
Solution Approach 1:
The patent segments the dialog handling into two distinct parts: a lightweight canned dialog system for routine multi-turn conversations and a sophisticated NLP system for complex intent identification. This segmentation allows each subsystem to operate independently at its optimal complexity level, resolving the contradiction between handling complexity and system simplicity.
Solution Approach 2:
The patent introduces an intermediary mechanism that routes user inputs between the canned dialog system and the NLP system based on the nature of the request. This intermediary allows the system to leverage both approaches appropriately, maintaining high accuracy for complex requests while avoiding unnecessary complexity for routine interactions.
2Reliability
If domain identification is performed for every user input, then the assistant can accurately operationalize user intent, but the processing time and computational resources increase
Solution Approach 1:
The patent segments dialog handling by routing routine multi-turn conversations through the lightweight canned dialog system without domain identification, while reserving domain identification for complex requests. This segmentation reduces processing time for the majority of interactions while maintaining accuracy when needed.
Solution Approach 2:
The patent applies partial domain identification by performing it only when necessary (for complex requests) rather than for every input. This partial action approach reduces overall processing time while maintaining sufficient accuracy for the system's needs.
3Productivity
If a lightweight predefined dialog scheme is used for handling user input, then deployment speed increases and computational resources are reduced, but the ability to handle complex natural language varies
Solution Approach 1:
The patent merges two previously separate systems (canned dialog and NLP) into a unified hybrid architecture. This combination allows the system to deploy quickly using the lightweight canned dialog framework while simultaneously maintaining the ability to handle complex natural language through the integrated NLP component.
Solution Approach 2:
The patent creates a universal dialog handling system that can perform both simple canned dialog responses and complex NLP-based intent identification through a single unified architecture. This multi-functionality allows the system to adapt to different request types without requiring separate deployment processes.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Systems and processes for providing multi-turn canned dialog are provided. An example method includes, at an electronic device, receiving a natural-language input; determining whether the natural-language input satisfies dialog criteria for multi-turn canned dialog; in accordance with a determination that the natural-language input satisfies the dialog criteria for multi-turn canned dialog: identifying a natural-language output of the multi-turn canned dialog corresponding to the natural-language input; and outputting the natural-language output; and in accordance with a determination that the natural-language input does not satisfy the dialog criteria for multi-turn canned dialog: identifying a task associated with the natural-language input; and performing the task associated with the natural-language input.