Knowledge Graph Refinement via Image Clustering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing knowledge graphs are often sparse and require manual input from subject matter experts to extend, leading to inefficiencies and informational gaps, such as missing objects, relationships, and attributes.

Innovation Solution

The method involves using image processing to identify scenes and objects within images, which are then clustered and used to refine knowledge graphs, reducing manual intervention and enhancing the breadth of information covered.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual input from subject matter experts is used to extend knowledge graphs, then the accuracy and reliability of knowledge graph content is improved, but the productivity and speed of knowledge graph extension deteriorates

Engineering Contradiction:

Improveaccuracy of knowledge graph contentVSAvoidspeed of knowledge graph extension

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent introduces image processing technology as an intermediary between raw visual data and the knowledge graph. The system automatically extracts objects, scenes, and relationships from images, serving as a mediator that bridges unstructured visual information and structured knowledge representations, thereby reducing manual expert intervention while maintaining knowledge quality

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the manual mechanical process of expert knowledge entry with automated image processing systems. Computer vision algorithms automatically identify and extract semantic information from images, substituting human experts' manual work with automated computational processes that scale more efficiently

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Loss of information

If manual input is used to add missing objects and relationships, then the completeness of knowledge graph information is improved, but the loss of time and computational resources worsens

Engineering Contradiction:

Improvecompleteness of knowledge graphVSAvoidtime for manual knowledge graph extension

Core Design Contradiction:

Loss of informationVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-processing images to extract objects, scenes, and relationships before integrating them into the knowledge graph. The system proactively identifies and structures information from image data in advance, preparing it for seamless integration and reducing the need for subsequent manual completion work

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system enables self-service by allowing the knowledge graph to automatically extend itself through image processing. The automated extraction and integration processes enable the knowledge graph to populate its own missing information without external manual intervention, making the system self-sufficient for knowledge expansion

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS10997231B2Image-based ontology refinement using clusters

Publication Date: 2021.05.04 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Aspects of the present disclosure relate to ontology refinement using image processing. A scene and a set of objects within the scene of an image are identified by performing image processing on the image. The set of objects is stored in a cluster labeled by the scene. A knowledge graph is then refined using the cluster.