Knowledge Graph Enrichment Using Targeted AI Property Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knowledge graph enrichment techniques are inefficient, costly, and prone to inaccuracies due to manual construction and the indiscriminate use of generative AI, leading to lengthy processes and unreliable data.
Innovation Solution
A method that generates tailored prompts for a language model based on identified relevant properties, limiting queries to those necessary for knowledge graph enrichment, thereby reducing computational costs and errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual construction techniques are used to enrich knowledge graphs, then reliability of information can be maintained, but the process becomes lengthy and tedious with low productivity
Solution Approach 1:
The patent introduces an intermediary system that acts as a bridge between manual construction and automated AI enrichment. This intermediary validates AI-generated properties against the knowledge graph schema and entity types, ensuring reliability while enabling faster automated processing. The intermediary layer filters and verifies information before integration, combining the speed of AI with the reliability of manual validation.
2Extent of automation
If generative AI systems are used to enrich knowledge graphs without preconception of desired properties, then automation is increased, but computational costs and energy consumption increase significantly
Solution Approach 1:
The patent applies preliminary action by first identifying which properties are actually needed for the entity enrichment based on the knowledge graph schema and entity type. This preliminary identification of required properties occurs before querying the generative AI system, ensuring that the AI only generates relevant information. This prevents wasteful computation on irrelevant properties and significantly reduces energy consumption while maintaining high automation.
3Quantity of substance
If generative AI systems query for all possible properties without direction, then comprehensive coverage is achieved, but the accuracy and consistency of enriched information decreases due to hallucinations
Solution Approach 1:
The patent applies local quality by tailoring the AI query to the specific entity type and required properties rather than using a generic comprehensive query. The system adapts the query structure, properties requested, and validation rules based on the local context of each entity being enriched. This ensures comprehensive coverage of relevant properties while maintaining high accuracy through context-aware querying and validation specific to each entity type.
4Quantity of substance
If numerous inferences are generated by the language model to cover all properties, then property coverage is increased, but the time required for enrichment increases and productivity decreases
Solution Approach 1:
The patent extracts only the necessary properties needed for entity enrichment based on the knowledge graph schema and entity type, rather than requesting all possible properties from the language model. This extraction of relevant properties before querying reduces the number of inferences needed, decreases enrichment time, and maintains comprehensive coverage of essential properties. The system takes out only what is needed, avoiding wasteful processing of unnecessary properties.
Data Source
Figure 1~2
Figure 3A~3B
Figure 4
AI summary
The invention relates to a method of communication with a computer device (CD) comprising a knowledge base (KB) modeling data in the form of a knowledge graph (KG), said method comprising the following, at the level of said device: - receiving (S1b) a first request (REQ1) comprising information relating to an entity (ENT) of the knowledge graph, - requesting (S2) the rendering of a web page containing said information, - receiving (S3b) a second request (REQ2) requesting at least one missing property of said entity, from among said rendered information, - requesting (S4) the rendering of a web page containing a list of missing properties ranked by frequency of observation of these properties for other entities of the same type, - querying (S5-S7) a language model, from a prompt generated in natural language by said device,said prompt asking the language model what the value (V) of at least one of the properties in the list is, - command (S8) the rendering of a web page containing said value.