Multi-Agent Dialogue Routing for Accurate Intent and Personalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional agents lack in-depth understanding of multi-turn dialogues and context, leading to inaccurate user intent recognition and poor personalized responses, resulting in a suboptimal user experience.
Innovation Solution
A multi-agent-based information processing method involving a first agent determining one or more second agents from a set based on input information, utilizing a combination of agents such as script generation, retrieval-augmented generation, proactive dialogue, and information search agents to enhance response accuracy and personalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a traditional single agent is used for customer service conversations, then the system structure is simple, but the user intent recognition accuracy is poor due to lack of multi-turn dialogue understanding
Solution Approach 1:
The patent divides the agent system into multiple specialized agents (first agent for intent recognition, second agent for response generation, third agent for personalization) to handle different aspects of customer service tasks. This segmentation allows each agent to focus on specific functions, improving overall intent recognition accuracy while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The patent introduces a temporal dimension by implementing multi-turn dialogue processing, where agents maintain conversation history and context across multiple interaction turns. This dimensional change from single-turn to multi-turn processing enables deeper understanding of user intent while the system manages complexity through structured dialogue state tracking.
2Adaptability or versatility
If a traditional agent provides responses without personalization, then the response generation process is simple, but the user experience is poor due to lack of personalized responses
Solution Approach 1:
The patent applies local quality by having a dedicated third agent responsible specifically for personalization tasks, while the second agent focuses on response generation. Each agent operates with specialized functionality tailored to its specific role, allowing the system to achieve high personalization capability without overwhelming complexity in the entire agent system.
Solution Approach 2:
The third agent acts as an intermediary between the second agent (response generation) and the final personalized response output. This intermediary component specifically processes user profile information and historical data to customize responses, adding personalization capability while isolating the complexity to a dedicated module rather than distributing it throughout the entire system.
3Reliability
If multiple agents are deployed to enhance dialogue understanding and personalization, then the response accuracy and user interaction effectiveness improve, but the system complexity increases
Solution Approach 1:
The patent segments the agent system into three distinct functional modules: first agent for intent recognition, second agent for response generation, and third agent for personalization. This segmentation improves reliability by ensuring each component performs its specific function expertly, while the modular structure makes the increased system complexity manageable and easier to maintain.
Solution Approach 2:
The patent implements a universal agent framework where each agent can potentially handle multiple related tasks within its domain. For example, the first agent handles intent recognition across various dialogue scenarios, and the third agent applies personalization to different response types. This multi-functionality approach improves response accuracy without proportionally increasing complexity, as agents are designed to be versatile within their functional scope.
Data Source
AI summary
A multi-agent-based information processing method includes: receiving an information processing request, in which the information processing request includes input information; inputting the input information to a first agent and obtaining output information of the first agent, in which the first agent determines one or more second agents from a set of agents based on the input information; and obtaining response information corresponding to the input information based on the output information of the first agent and the one or more second agents.