Fairness and Explanations in Entity Resolution: An Overview

Entity Resolution (ER) is a fundamental task in data integration, enabling the identification of records that refer to the same real-world entity across diverse and often heterogeneous data sources. Recent advances in Artificial Intelligence (AI) have significantly improved ER performance, particula...

Full description

Saved in:
Bibliographic Details
Main Authors: Tiago Brasileiro Araujo, Vasilis Efthymiou, Kostas Stefanidis
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11129015/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Entity Resolution (ER) is a fundamental task in data integration, enabling the identification of records that refer to the same real-world entity across diverse and often heterogeneous data sources. Recent advances in Artificial Intelligence (AI) have significantly improved ER performance, particularly with deep learning and pre-trained embeddings. However, these AI-driven solutions introduce new challenges related to fairness and explainability. Fairness-aware ER seeks to mitigate bias that may arise from algorithmic decision-making or imbalanced training data, while eXplainable Entity Resolution (XER) aims to enhance transparency and trust in ER. In this work, we provide a comprehensive overview of fairness and explainability in ER, systematically analyzing existing techniques across the ER pipeline. We discuss challenges in ensuring unbiased and interpretable ER outcomes, with a special focus on streaming environments, where real-time decision-making intensifies the complexity of these concerns. Furthermore, we outline research opportunities and examine the trade-offs between schema-aware and schema-agnostic methods, as well as rule-based and machine learning-based comparison techniques, in ensuring fair and transparent ER. Our study highlights open research challenges and potential future directions, encouraging novel explainable AI methodologies and fairness-aware ER solutions that enhance the reliability, accountability, and societal impact of AI-driven ER systems.
ISSN:2169-3536