Conversational AI Memory Graph for Personalized Multi-Turn Dialogue
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI virtual assistants lack the ability to personalize conversations, follow medical guidelines, and handle multi-turn conversations, leading to inefficiencies and compliance with healthcare regulations, and often fail to contextualize interactions, and lack the capability to contextualize interactions, and are not designed for use in healthcare settings.
Innovation Solution
Implement a conversational AI system that includes a large language model (LLM) configured to execute and store a multi-turn conversation with a user, in a memory device, where a conversation analyzer processes a preceding multi-turn conversation to detect and store at least some turns in the multi-turn conversation that deviate from a target topic and instead characterize the life attributes of the user, and construct a knowledge graph based on the life attributes of the user and store the knowledge in a tokenized format, for subsequent use in a subsequent multi-turn conversation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a conventional AI virtual assistant is used, then the system is simple to operate, but it cannot personalize conversations or contextualize interactions
Solution Approach 1:
The system segments the conversation processing into distinct functional modules: a conversation executor that handles the primary medical topic discussion, a conversation analyzer that identifies life attribute information, and a knowledge graph applicator that personalizes responses. This segmentation allows each component to specialize in specific tasks, enabling personalization without requiring the entire system to be fundamentally complex.
Solution Approach 2:
The patent introduces a knowledge graph as an intermediary data structure that stores and organizes user life attributes. This knowledge graph acts as a mediator between the conversation analyzer (which extracts attributes) and the LLM (which uses attributes for personalization), allowing the system to maintain simplicity while achieving adaptability through this intermediate layer.
2Loss of information
If the AI analyzes and stores all conversation turns, then complete user information is captured, but processing time and computational resources increase
Solution Approach 1:
The conversation analyzer selectively extracts only the relevant life attribute information from conversation turns that deviate from the target topic, rather than processing and storing all conversation data. This extraction approach captures essential personalization information while minimizing processing overhead and storage requirements.
Solution Approach 2:
The system performs preliminary analysis during the conversation to identify and store life attribute information as it becomes relevant, rather than requiring a comprehensive post-conversation analysis. This allows the system to progressively build the knowledge graph during natural interaction, reducing overall processing time.
3Reliability
If the AI follows a strict medical checklist, then compliance with medical guidelines is ensured, but the conversation becomes less natural and personalized
Solution Approach 1:
The system dynamically adjusts the conversation flow by allowing deviations from the strict medical checklist when life attribute information is detected. The LLM can naturally incorporate personalization elements while the underlying medical checklist remains as a reliability guide, creating a dynamic balance between compliance and naturalness.
Solution Approach 2:
The patent applies different qualities to different parts of the conversation: the core medical information exchange follows strict checklist guidelines for reliability, while peripheral conversational elements can be personalized using life attributes. This local differentiation allows medical compliance in critical areas while maintaining naturalness in social aspects.
4Productivity
If the AI stores and retrieves knowledge graphs in tokenized format, then information retrieval efficiency improves, but memory storage requirements increase
Solution Approach 1:
The system changes the parameter of knowledge graph storage from full-structure format to tokenized format, which compresses the representation while maintaining essential information. This parameter change improves retrieval efficiency through faster token-based access while reducing the storage burden compared to storing complete knowledge graph structures.
Data Source
AI summary
A conversational artificial intelligence (AI) system is configured to engage in a multi-turn conversation with a user. The multi-turn conversation is substantially focused on a target topic. A conversation analyzer analyzes the multi-turn conversation to detect and store at least some turns in the multi-turn conversation that deviate from the target topic and instead characterize life attributes of the user. A knowledge graph constructor builds a knowledge graph for the user based on at least some turns in the multi-turn conversation that characterize the life attributes of the user. The knowledge graph translates the life attributes of the user into the user's life biography, including life chronology, life preferences, life milestones, life events, or any combination thereof. A knowledge graph applicator uses parts of the knowledge graph in a subsequent multi-turn conversation with the user by contextually interspersing portions of the user's life biography in the subsequent multi-turn conversation.