Ontology-Based Entity Resolution Using ML Feedback

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing entity resolution processes face challenges in efficiently matching and resolving records across multiple data sources, particularly due to the large volume of records and the need for accurate identification of real-world entities.

Innovation Solution

A system is developed that uses machine learning techniques to determine whether incoming records correspond to known entities within an ontology framework, allowing for customizable entity resolution operations and feedback-driven improvement of machine learning models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional record matching methods are used to compare records across multiple data sources, then the process can identify potential matches, but the large volume of records causes inefficiency and high computational overhead

Engineering Contradiction:
Improverecord matching efficiencyVSAvoidprocessing time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent segments the entity resolution process into distinct stages: blocking (grouping records by common attributes), comparison (detailed record pairing), and clustering (grouping matches). This segmentation reduces the computational complexity from O(n²) to O(n) by first grouping records into manageable blocks before performing detailed comparisons, thereby improving efficiency and reducing processing time for large volumes of records

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action through the blocking stage, where records are pre-grouped by common attributes (such as name, location, or time) before the actual comparison process. This preliminary organization eliminates the need to compare every record against every other record, significantly reducing the search space and processing time required for entity resolution

Inventive Principle:
Principle #10Preliminary action

2Productivity

If automated machine learning methods are used for entity resolution, then processing speed increases, but accuracy and reliability of match determination decrease

Engineering Contradiction:
Improveprocessing speedVSAvoidmatch accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements feedback mechanisms where user corrections and confirmations of match decisions are captured and used to retrain the machine learning models. This continuous feedback loop allows the system to learn from errors and improve accuracy over time, maintaining high reliability while preserving automated processing speed. The system adapts to domain-specific nuances through this iterative learning process

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent employs dynamic adjustment of matching thresholds and model parameters based on confidence scores. When the machine learning model is highly confident about a match, automated determination proceeds quickly; when confidence is lower, the system adjusts thresholds or requests additional verification. This dynamic approach maintains processing speed while ensuring accuracy for uncertain cases

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If detailed comparison methods are applied to all record pairs, then match accuracy improves, but the computational complexity and resource requirements increase significantly

Engineering Contradiction:
Improvematch determination accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by performing detailed comparisons only on specific fields and attributes that are most relevant to entity identification, rather than uniformly analyzing all record attributes. The system dynamically selects which fields to compare based on the blocking group and record characteristics, reducing computational complexity while maintaining accuracy for critical matching attributes

Inventive Principle:
Principle #3Local quality

4Reliability

If custom entity resolution operations are implemented to handle specific domain requirements, then resolution accuracy for domain-specific entities improves, but the system becomes less adaptable to different domains

Engineering Contradiction:
Improvedomain-specific resolution accuracyVSAvoidsystem flexibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent implements a universal ontology-based framework that can represent multiple entity types (persons, organizations, locations, etc.) and domain-specific attributes through a common structure. The machine learning models are trained on domain-specific data but operate within this universal framework, allowing the system to adapt to different domains by configuring the ontology appropriately rather than requiring complete system redesign. This multi-functionality enables both domain-specific accuracy and cross-domain versatility

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4560492A1Entity resolution based on ontology structures
Publication Date: 2025.05.28 PALANTIR TECHNOLOGIES INC
  • EP4560492A1 patent drawingFigure 1
  • EP4560492A1 patent drawingFigure 2
  • EP4560492A1 patent drawingFigure 3

AI summary

A system for managing entity resolution processes is disclosed. The system is programmed to determine whether incoming records correspond to known entities within an ontology framework. The system is also programmed to manage a graphical user interface (GUI) that allows customizing entity resolution operations and providing feedback on the determination results. The system is further programmed to use the provided feedback to improve machine learning for the entity resolution processes.