Pre-trained language models accurately link records and infer ethnicity from names
Pre-Trained Language Models as Zero-Shot Tools for Social Science Research
Computation and Language
Summary
Many social science tasks need to match records or guess people’s ethnicity using their names, but getting good training data for this is hard. The paper shows that large language models, which are already trained on lots of text, can do these tasks well without extra training. They introduce a method called EnsembleLink for matching records that works better than many past approaches, even across different languages and formats. Also, these models can guess ethnicity more accurately than some common techniques and can handle groups beyond the U.S. This means the knowledge learned by the models is often enough for these tasks without needing specific extra examples.
What this means in practice
- •For social data analysts: Improve matching of records from diverse sources and formats without extra training data by using EnsembleLink with pre-trained embeddings.
- •For demographic research teams: Enhance ethnicity inference accuracy from names, including beyond U.S. groups, using large language models without custom models or labels.
Authors
Noah Dasanaike
Abstract
Record linkage and demographic inference from names are routine demands in social science. Most approaches to these problems require labeled training data that are difficult or costly to acquire. I show that pre-trained language models perform both tasks accurately without requiring task-specific training. For record linkage, I introduce EnsembleLink, an algorithm that uses pre-trained embeddings alongside string-based scorers to retrieve a short list of candidates and then identify the likeliest matches. EnsembleLink matches or exceeds the performance of most prior approaches while generalizing across scripts and formats that those methods are unable to handle. For ethnicity inference from names, I show that large language models (LLMs) achieve up to 83.8% accuracy on U.S. voter files, compared to 78.0% for name-supplemented BISG. Open-weight models offer reproducible implementations that approach these models in accuracy. Moreover, LLMs are able to infer group membership for religious, caste-based, and other ethnic group categories outside of the United States, and reduce bias in downstream estimates of geographic racial gaps. Where pre-training encodes the world knowledge required by each task, task-specific labels add less information.