Query Rewriting via Knowledge Graph for Contextual Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Natural language processing systems struggle to provide contextually relevant responses to ambiguous queries or when information is unavailable, often generalizing questions too much and resulting in false hits.
Innovation Solution
A method that interprets user queries using schema information, rewrites them into related queries, prunes irrelevant ones based on contemporary circumstances, filters by user preferences, and uses a knowledge graph to provide answers, thereby enhancing contextual accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If natural language processing systems generalize ambiguous queries to provide responses, then they can maintain interaction flow, but the accuracy and contextual relevance of responses deteriorates resulting in false hits
Solution Approach 1:
The system segments the query processing into multiple stages: initial query interpretation, schema information matching, knowledge graph traversal, and preference-based filtering. This segmentation allows the system to handle ambiguous queries systematically rather than through simple generalization, maintaining both productivity and precision.
Solution Approach 2:
The patent introduces an intermediary knowledge graph that acts as a mediator between the user's ambiguous query and the final response. The knowledge graph contains structured information about the querent and contextual data, allowing the system to resolve ambiguities without over-generalizing, thus maintaining contextual accuracy while enabling interaction.
2Measurement precision
If the system retrieves and processes multiple related queries to improve answer accuracy, then the quality of responses improves, but the processing time and system complexity increases
Solution Approach 1:
The system performs preliminary actions by pre-building and maintaining a knowledge graph with structured information about users, their preferences, and contextual data. This preliminary structuring allows the system to efficiently process related queries without ad-hoc complexity, as the knowledge graph is already organized for rapid traversal and matching.
Solution Approach 2:
The system changes parameters by dynamically adjusting the scope of query processing based on confidence levels and preference weights. When initial queries yield low-confidence results, the system selectively expands to related queries only when necessary, controlled by parameters such as preference weights and confidence thresholds, thereby managing complexity while maintaining accuracy.
3Reliability
If the system uses schema information and knowledge graphs to interpret queries accurately, then contextual relevance improves, but the computational resources and processing time required increases
Solution Approach 1:
The system applies local quality by focusing computational resources on specific relevant portions of the knowledge graph rather than processing the entire structure. The preference-based query filtering mechanism identifies and processes only those related queries that are locally relevant to the user's preferences and the original query context, reducing overall computational resource usage while maintaining high contextual relevance.
Data Source
AI summary
According to one embodiment, a method, computer system, and computer program product for ancillary speech generation is provided. The present invention may include receiving a query from a querent; interpreting the terms of the query using schema information; rewriting the initial query into a set of related queries; pruning related queries that are irrelevant based on contemporary circumstances; retrieving preferences associated with the querent, where each of the preferences is assigned a weight representing the importance of that preference to the querent; filtering out the pruned related queries that do not comport with the querent's preferences; sort the filtered related queries according to the weight of the preferences served by each query; answering the filtered related queries with a knowledge graph; returning answers to the querent; and updating the preferences based on the answers.