Entity Resolution Rules via Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity resolution methods face challenges in accurately matching entities due to the use of predefined similarity functions that may not be accurate or sufficient across various domains, and the lack of interpretability in black box classifier models.
Innovation Solution
The use of deep neural networks to learn entity resolution rules from historical pairs of entities, developing an entity resolution model that includes sets of rules and conditions, and applying this model to new pairs of entities to determine matching or non-matching status.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If predefined similarity functions are used for entity resolution, then the implementation is simple, but the matching accuracy is insufficient across various domains
Solution Approach 1:
The patent transforms the entity resolution approach by changing from fixed predefined similarity functions to dynamic learned parameters through deep neural networks. The system learns domain-specific matching parameters and rules from historical data, allowing the matching criteria to adapt to different domains while maintaining implementation simplicity through automated learning.
Solution Approach 2:
The system implements self-service by automatically learning entity resolution rules from historical paired entity data without requiring manual configuration of similarity functions. The deep neural network autonomously identifies matching patterns and generates resolution rules, eliminating the need for domain experts to manually tune parameters while achieving high accuracy.
2Measurement precision
If black box classifier models are used for entity resolution, then the matching accuracy can be improved, but the interpretability is lost
Solution Approach 1:
The patent segments the black box classifier into interpretable components by extracting explicit resolution rules from the deep neural network's learned parameters. Instead of presenting a single opaque classification result, the system breaks down the decision-making process into discrete, understandable rules that can be independently analyzed and explained, maintaining both accuracy and interpretability.
Solution Approach 2:
The patent introduces an intermediary layer between the deep neural network and the final classification output. This intermediary extracts and formulates the learned parameters as human-readable resolution rules, serving as a mediator that translates the network's internal representations into interpretable forms without sacrificing the accuracy benefits of the neural network approach.
Data Source
AI summary
A method, system, and computer program product for learning entity resolution rules for determining whether entities are matching. The method may include receiving historical pairs of entities. The method may also include determining a set of rules for determining whether a pair of entities are matching, where the set of rules comprises a plurality of conditions. The method may also include developing, using a deep neural network, an entity resolution model based on the historical pairs of entities. The method may also include receiving a new pair of entities. The method may also include applying the entity resolution model to the new pair of entities. The method may also include determining whether one or more rules from the set of rules are satisfied for the new pair of entities. The method may also include categorizing the new pair of entities as matching or not matching.