Entity Matching via Phonetic Blocking and Transitive Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to identify entities involved in multiple transactions, particularly in financial transactions, where records may differ significantly, leading to challenges in accurately linking customer information across various databases and products.
Innovation Solution
The system employs a method that includes accessing multiple databases containing transaction records, standardizing and comparing these records using metaphone algorithms to approximate sound similarities, and applying blocking criteria to limit comparisons, ultimately generating probabilities of entity matches and transitive linking to identify groups of records involving the same entity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional record comparison methods are used, then exact matches can be identified, but entities with similar but different records cannot be linked
Solution Approach 1:
The system transforms string-based record comparisons into phonetic-based comparisons by converting names and addresses into their phonetic representations. This parameter change enables the system to match records that are phonetically similar but spelled differently, thereby improving entity identification accuracy while adapting to various record formats.
Solution Approach 2:
The system introduces phonetic codes as an intermediary representation between the original record data and the matching process. By converting names and addresses into phonetic forms, the system creates a common representation space that facilitates accurate matching across different record formats and spelling variations.
2Reliability
If all records are compared to ensure comprehensive matching, then complete entity identification is achieved, but computational resources and time are exhausted
Solution Approach 1:
The system segments the record comparison process into multiple stages: first grouping records by phonetic codes, then comparing only within phonetic groups. This segmentation dramatically reduces the number of comparisons needed while maintaining comprehensive entity identification, as records with different phonetic codes cannot match anyway.
Solution Approach 2:
The system performs preliminary phonetic code generation and grouping before the actual comparison process. By pre-computing phonetic representations and organizing records into groups based on these codes, the system prepares the data structure in advance, enabling fast and efficient matching without requiring exhaustive comparisons.
3Productivity
If phonetic codes and blocking criteria are applied, then computational efficiency is improved, but matching precision may be reduced
Solution Approach 1:
The system uses a feedback mechanism where the initial phonetic code-based matching is followed by a verification step that compares the actual record fields (name, address, date of birth) to confirm the match. This feedback loop ensures that phonetic similarities are validated with actual data, maintaining high match accuracy while benefiting from the efficiency gains of phonetic grouping.
Data Source
AI summary
There are provided systems and methods for determining an entity involved in multiple transactions. The system may include at least one database containing transaction records. It further may include a control circuit that: receives data; creates a metaphone of entries in the records; creates a number of record pairs; compares the two records of each record pair a comparison value indicating a degree of similarity of the two records; and generates a probability that the two records involve the same entity. The control circuit may also iteratively apply different fields and variables in the records to determine that each record pair is either a match or a non-match and then link matched record pairs with other corresponding matched record pairs to determine an entity.