Multi-round Session Interaction Using Requirement Structure Trees
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-round session interaction systems are suboptimal for task-based sessions with tree-like hierarchical dependency relationships, as they are complex to customize and prone to conflicts, and statistical dialogue systems face challenges in acquiring and understanding training corpora.
Innovation Solution
A method and system that map user requirement information to standard requirements in locally preset requirement structure trees, allowing for the output of content from sub-nodes or session information if the standard requirement matches the user input, improving the matching rate and handling tree-like hierarchical dependencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a multi-round session interaction system defines rules artificially to customize a task, then the system can handle specific tasks, but the system becomes complex and prone to conflicts between rules
Solution Approach 1:
The patent introduces a requirement structure tree as an intermediary representation layer between user natural language and system actions. The tree maps user requirements to standardized nodes, which then trigger corresponding system actions. This intermediary structure simplifies the interaction by providing a structured, conflict-free mapping mechanism rather than relying on multiple artificial rules
Solution Approach 2:
The patent segments the session interaction system into distinct modular components: requirement structure trees, node mapping mechanisms, and action execution modules. Each requirement tree is independently structured with hierarchical nodes, allowing the system to handle complex tasks through composition of simpler, well-defined segments rather than monolithic rule sets
2Extent of automation
If a statistical dialogue system uses reinforcement learning to learn tree-like dependency relationships, then the system can automatically learn relationships, but the training corpus is difficult to acquire and the learned content is hard to understand and control
Solution Approach 1:
The patent performs preliminary structuring of domain knowledge into requirement structure trees before actual dialogue processing. These trees are pre-defined with standardized nodes and hierarchical relationships, eliminating the need for complex training corpus acquisition and reinforcement learning. The structure is prepared in advance to guide the interaction flow
Solution Approach 2:
The requirement structure tree serves as an interpretable intermediary that bridges user input and system actions in a transparent manner. Unlike black-box reinforcement learning models, the tree structure provides explicit, human-understandable mappings between requirements and actions, allowing for easy analysis and control of the learned content
3Productivity
If existing systems use pre-defined system state and action set with strategies or statistical models, then the system can operate automatically, but the system is suboptimal for task-based sessions with tree-like hierarchical dependency relationships
Solution Approach 1:
The patent transitions from flat, single-dimensional state-action mappings to a hierarchical tree structure with multiple dimensions. The requirement structure trees introduce hierarchical depth, allowing the system to naturally represent and process tree-like dependency relationships that cannot be effectively captured in traditional flat state-action frameworks
Data Source
AI summary
A method for interacting in a multi-round session includes: receiving user requirement information; mapping the user requirement information to standard requirements contained in nodes of a plurality of locally preset requirement structure trees, the standard requirement contained in each node being a general requirement of standard requirements contained in sub-nodes of the node in the requirement structure trees; and outputting content of the standard requirements contained in sub-nodes of a selected non-leaf node or session information corresponding to a selected leaf node, if the standard requirement contained in the selected node in the plurality of requirement structure trees matches the user requirement information.