Knowledge Graph Query Optimization via Reinforced Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine-learned approaches for conversational programs, such as bots and personal assistants, lack grounding in knowledge graphs, leading to inefficient user query understanding and task completion, as they rely on similarity-based responses rather than contextually relevant interactions.
Innovation Solution
The integration of a reinforced learning mechanism with a knowledge-graph simulation score to determine optimal response queries, which narrows down user interactions by identifying user interest and filtering options, thereby reducing the number of queries needed to complete a task accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine-learned models use similarity-based responses without knowledge graph grounding, then the system complexity remains low, but the understanding accuracy of user queries deteriorates
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary between the user query and the response generation system. The knowledge graph stores structured information about entities, attributes, and relationships, allowing the bot to ground its responses in factual knowledge rather than relying solely on similarity-based pattern matching. This intermediary structure enables more accurate understanding of user queries by providing contextual knowledge about the entities being discussed.
Solution Approach 2:
The system segments the query understanding process into multiple components: parsing the user query to identify entities and attributes, querying the knowledge graph for relevant information, and generating responses based on the retrieved knowledge. This segmentation allows each component to specialize in a specific task, improving overall understanding accuracy while managing system complexity through modular design.
2Manufacturing precision
If the bot asks more questions to clarify user intent, then the task completion accuracy improves, but the number of conversation turns increases
Solution Approach 1:
The knowledge graph performs preliminary action by pre-storing structured information about entities, their attributes, and relationships. When a user query is received, the system can immediately query the knowledge graph for relevant information without needing to ask multiple clarifying questions. For example, if the user mentions a product, the knowledge graph already contains its attributes, allowing the system to provide accurate responses in fewer conversation turns.
Solution Approach 2:
The system uses feedback from the knowledge graph query results to determine whether additional clarifying questions are needed. If the knowledge graph returns sufficient information to address the user's query, the system can proceed directly to response generation. If the results are insufficient, the system selectively asks targeted questions to obtain missing information, thereby minimizing the number of conversation turns while maintaining high task completion accuracy.
3Stability of the object's composition
If the bot uses rule-based approaches for conversation, then the consistency of responses improves, but the adaptability to different user queries deteriorates
Solution Approach 1:
The knowledge graph serves as a universal knowledge base that can support multiple types of queries across different domains. Rather than requiring separate rule sets for different query types, the system queries the knowledge graph for relevant information and generates appropriate responses based on the retrieved data. This multi-functional approach maintains response consistency through structured knowledge while adapting to diverse user queries.
Data Source
AI summary
The technology described herein allows an interactive program to leverage a knowledge graph to maximize the likelihood of successfully understanding the user's query and at the same time minimize the number of turns taken to understand the user. A goal of the technology described herein is to formulate response queries that have a probability of completing the user's requested task accurately while issuing the fewest number of response queries to the user before determining the intended task. In order to accomplish this, the technology combines a reinforced learning mechanism with a knowledge-graph simulation score to determine the optimal response query.