Dialogue Graph Context Routing for AI Task Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AI models face challenges in efficiently processing large datasets on-device due to high operational costs and time requirements, limiting their use in task-oriented dialogue systems.

Innovation Solution

A method and system that utilize a dialogue graph and dialogue model to determine the context of a task-oriented dialogue, select the appropriate task performing model, and execute the task, optimizing resource usage and improving efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a mixture of experts (MoE) architecture is utilized to improve AI model efficiency, then calculation resources are concentrated and overall efficiency is improved, but high-level video random access memory (VRAM) is required and the fine tuning process becomes complex

Engineering Contradiction:
ImproveAI model processing efficiencyVSAvoidmodel architecture complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the AI model into multiple expert models, each specialized in handling specific types of tasks or contexts. The gating network segments the input data and routes it to the appropriate expert model, allowing the system to process different tasks using only the necessary computational resources rather than activating the entire model.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic model architecture where the system can adaptively activate only the required expert models based on the specific task at hand. This dynamic activation pattern allows the system to optimize resource usage in real-time, reducing memory requirements and computational overhead while maintaining high processing efficiency for specific task types.

Inventive Principle:
Principle #15Dynamics

2Use of energy by moving object

If a mixture of experts (MoE) architecture is utilized to manage large size models, then resource efficiency is improved for activated tasks, but support for non-activated resources becomes limited

Engineering Contradiction:
Improvecomputational resource efficiencyVSAvoidresource allocation flexibility
Core Design Contradiction:
Use of energy by moving objectVSAdaptability or versatility

Solution Approach 1:

The patent designs the gating network to universally handle various types of input data and dynamically select from multiple expert models, making the system adaptable to different task types. Each expert model can be independently trained and updated, allowing the system to maintain versatility while optimizing resource usage for each specific task domain.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The gating network serves as an intermediary between the input data and the expert models, intelligently routing data to the most appropriate expert based on the task requirements. This intermediary layer provides flexible resource allocation by dynamically determining which expert models should be activated, thereby balancing computational efficiency with adaptability across diverse task types.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If AI models are implemented to process large amounts of data for task-oriented dialogues, then service capability is improved, but operational cost and processing time become considerable

Engineering Contradiction:
Improvetask processing capabilityVSAvoiddata processing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent segments the dialogue processing task into different expert models, each specialized in handling specific dialogue acts or task types. This segmentation allows the system to process dialogue data more efficiently by routing it to the appropriate expert model, reducing overall processing time while maintaining comprehensive task处理能力.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements pre-training for each expert model on specific task types and contexts before deployment. This preliminary action allows the expert models to be pre-equipped with task-specific knowledge and patterns, enabling them to process incoming dialogue data more quickly and efficiently without requiring extensive real-time computation for learning.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4567629A1Method and system for performing tasks based on the context of task-oriented dialogue
Publication Date: 2025.06.11 LG MANAGEMENT DEV INST CO LTD
  • EP4567629A1 patent drawingFigure 1
  • EP4567629A1 patent drawingFigure 2
  • EP4567629A1 patent drawingFigure 3

AI summary

A task performing method may perform a task based on a context of a task-oriented dialogue through a dialogue model by a computing system including a memory and a processor. The task performing method may include: receiving a user dialogue input; determining and providing a response dialogue act to the user dialogue input based on a dialogue graph; determining the context of the task-oriented dialogue by analyzing data of a series of task-oriented dialogues including the user dialogue input and the response dialogue act; determining the type of task requested by a user based on the context of the task-oriented dialogue; and performing the task of which type is determined.