Contextual Knowledge Graph Cleansing for Accurate Error Review

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing data cleansing methods for knowledge graphs are inefficient and prone to human error, requiring manual review of vast datasets and lacking verification mechanisms, leading to inaccurate and incomplete data.

Innovation Solution

A method for contextual visualization and cleansing of knowledge graphs using a data processing device that segregates, displays, and verifies node data based on user context, employing techniques like natural language processing and machine learning to identify and correct errors automatically.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional data cleansing techniques are used to identify and rectify errors in knowledge graphs, then data accuracy can be improved, but the process requires extensive human effort and time to review vast datasets

Engineering Contradiction:
Improvedata accuracyVSAvoidtime for data review
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and displays only contextually relevant data points from the knowledge graph based on user context, rather than presenting all data points. This selective extraction allows users to focus on relevant errors without being overwhelmed by the entire dataset, thereby improving data accuracy review while reducing time loss.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system changes the parameter of data presentation from showing all data points to showing only contextually relevant data points. This parameter change in data filtering and presentation reduces the volume of data requiring human review while maintaining the ability to identify errors accurately.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If all data points of the knowledge graph are presented to the user for review, then complete data coverage is achieved, but the user must examine nearly impossible volumes of data to identify errors

Engineering Contradiction:
Improvedata coverageVSAvoidease of data review
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The system extracts and presents only contextually relevant data points based on user context, rather than displaying all data points. This maintains reliable data coverage for the user's specific needs while dramatically improving ease of operation by reducing the review burden.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of presenting all data points (excessive action), the system presents only the necessary subset of contextually relevant data points (partial action). This partial presentation is sufficient for error identification while making the task operable.

Inventive Principle:
Principle #16Partial or excessive action

3Ease of manufacture

If human users manually review and correct data errors in knowledge graphs, then data cleansing can be performed, but human errors may occur and lack verification mechanisms

Engineering Contradiction:
Improvedata cleansing capabilityVSAvoiderror verification
Core Design Contradiction:
Ease of manufactureVSReliability

Solution Approach 1:

The patent implements a verification mechanism where corrected data points are verified before final acceptance. This feedback loop allows the system to check and validate user corrections, reducing human errors and improving reliability of the data cleansing process.

Inventive Principle:
Principle #23Feedback

4Measurement precision

If traditional data cleansing methods are used, then errors can be identified, but the process is inefficient and requires expert human intervention for large volumes of data

Engineering Contradiction:
Improveerror identification accuracyVSAvoiddata cleansing efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system extracts and presents only contextually relevant data points for review, dramatically improving productivity by reducing the volume of data requiring human examination while maintaining error identification accuracy through contextual filtering.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system performs partial data presentation showing only necessary data points for error identification, which significantly improves cleansing efficiency while maintaining the precision of error detection through contextual relevance filtering.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4703901A1Method for contextual visualization and cleansing of a knowledge graph
Publication Date: 2026.03.04 SIEMENS AG
  • EP4703901A1 patent drawingFigure 1
  • EP4703901A1 patent drawingFigure 2
  • EP4703901A1 patent drawingFigure 3

AI summary

The method comprises receiving an original knowledge graph (100) comprising nodes (N1 to Nn), and edges (E 1 to En) that define relationship between the nodes. Further, the node data associated with each of the nodes is segregated and a contextual node data associated with the node is determined based on a query received from a user. Further, the contextual node data is displayed to the user. Further, a verification operation is performed on the original knowledge graph (100) to identify errors. Further, one or more inputs is received from the user for updating the original knowledge graph (100) to correct the identified errors. Further, the original knowledge graph (100) is updated based on the inputs received from the user to obtain an updated knowledge graph.