Adaptive Record Linking via Dynamic Matching Rules

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing record linking systems face challenges in matching incomplete, erroneous, or non-standard data records due to rigid matching rules, leading to inefficiencies in recall, speed, and precision when searching for candidate records.

Innovation Solution

An adaptive learning machine is employed to update database records with aliases such as phonetic spellings, initials, and anagrams, and uses information retrieval heuristics to match and rank records based on similarity, eliminating the need for user-maintained matching rules.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If rigid matching rules are used to match data records, then matching precision can be maintained, but recall and adaptability deteriorate due to inability to handle incomplete, erroneous, or non-standard data

Engineering Contradiction:
Improvematching precisionVSAvoidadaptability to incomplete or erroneous data
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamic matching rules that adapt based on data quality and context. The system adjusts matching thresholds and criteria dynamically rather than using fixed rigid rules, allowing it to handle incomplete, erroneous, or non-standard data while maintaining precision through context-aware adjustments.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes matching parameters adaptively based on the characteristics of the data being processed. When encountering incomplete or erroneous records, the system modifies matching thresholds, weightings, and criteria to accommodate varying data qualities while still achieving accurate matches.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If comprehensive matching criteria are applied to handle fuzzy matching, then recall improves, but processing speed deteriorates

Engineering Contradiction:
ImproverecallVSAvoidprocessing speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the matching process into multiple stages: initial filtering using quick criteria to eliminate obviously mismatched records, followed by more comprehensive evaluation only for promising candidates. This segmentation maintains high recall while improving processing speed by avoiding exhaustive evaluation of all records.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system applies partial matching criteria initially to quickly identify candidate records, then applies full comprehensive criteria only to those candidates. This partial action approach achieves high recall for the final result set while maintaining processing speed by limiting exhaustive evaluation to a smaller subset.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If user-maintained matching rules are used, then matching precision can be controlled, but system complexity and maintenance cost increase

Engineering Contradiction:
Improvematching precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements self-service matching where the system automatically learns and adjusts matching rules based on observed data patterns and user feedback, eliminating the need for manual rule creation and maintenance. The system serves itself by continuously improving its matching algorithms through adaptive learning while maintaining precision.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system incorporates feedback mechanisms where matching results and user corrections are fed back into the rule generation process. This automatic feedback loop allows the system to refine its matching rules continuously without manual intervention, reducing complexity while maintaining or improving precision over time.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS9552393B2Adaptive record linking in a distributed computing system
Publication Date: 2017.01.24 SAP IRELAND LTD
  • US9552393B2 patent drawing
  • US9552393B2 patent drawing
  • US9552393B2 patent drawing

AI summary

Techniques for information retrieval include the features of receiving a plurality of data records, updating a plurality of database records associated with the received plurality of data records stored in a database, receiving a query for a particular database record, and preparing for display, in response to the query, one or more of the database records based on at least one of the name of the business enterprise or the alias associated with the name of the business enterprise. Each data record includes data fields associated with a business enterprise. The data fields include a name of the business enterprise. Each database record includes attributes including the name of the business enterprise and an alias associated with the name of the business enterprise. The query includes at least one of the name of the business enterprise or the alias associated with the name of the business enterprise.