Code accompanying the paper: "KAN-ER: Kolmogorov-Arnold Networks-Based Entity Resolution in Data Lakes" DaWaK 2026
Entity Resolution (ER) is a fundamental task in data integration and data lakes, aiming to identify records that refer to the same real-world entity across heterogeneous data sources.
This project introduces KAN-ER, a lightweight and interpretable Entity Resolution framework based on Kolmogorov-Arnold Networks (KAN).
The framework operates on attribute-level similarity vectors, learns nonlinear matching functions, and provides interpretable spline-based reasoning for attribute contributions.
The evaluation includes:
- Structured records
- Noisy (dirty) records
- Textual record pairs
Across several domains, including:
- Music
- Bibliography
- E-commerce
Python 3.9+ is recommended.
Install the required dependencies: pip install -r requirements.txt
The benchmark datasets used in this work can be downloaded from the Magellan project:
https://github.com/anhaidgroup/deepmatcher/blob/master/Datasets.md
This project is licensed under the MIT License. See the LICENSE file for details.
This work was carried out with the support of the ANRT through the CIFRE (Industrial Agreements for Training through Research) program.
The project is supported by:
- Trimane
- LIAS laboratory (ISAE-ENSMA)