EnsembleLink: Accurate Record Linkage Without Training Data

TL;DR

EnsembleLink is a novel record linkage method that achieves high accuracy without training data by leveraging pre-trained language models, outperforming existing approaches on various benchmarks.

Contribution

It introduces EnsembleLink, a training-free record linkage approach using pre-trained language models, improving accuracy and efficiency over traditional methods.

Findings

Matches or exceeds accuracy of label-dependent methods

Operates locally without external API calls

Completes linkage tasks in minutes

Abstract

Record linkage, the process of matching records that refer to the same entity across datasets, is essential to empirical social science but remains methodologically underdeveloped. Researchers treat it as a preprocessing step, applying ad hoc rules without quantifying the uncertainty that linkage errors introduce into downstream analyses. Existing methods either achieve low accuracy or require substantial labeled training data. I present EnsembleLink, a method that achieves high accuracy without any training labels. EnsembleLink leverages pre-trained language models that have learned semantic relationships (e.g., that "South Ozone Park" is a neighborhood in "New York City" or that "Lutte ouvriere" refers to the Trotskyist "Workers' Struggle" party) from large text corpora. On benchmarks spanning city names, person names, organizations, multilingual political parties, and bibliographic records, EnsembleLink matches or exceeds methods requiring extensive labeling. The method runs locally on open-source models, requiring no external API calls, and completes typical linkage tasks in minutes.