Name Ethnicity Detector

Overview

This toolkit predicts likely ethnicity labels from names using character-level modeling pipelines and multiple model configurations. It is oriented toward research and audit use-cases where direct demographic labels are unavailable.

Data and Sources

• Input CSV workflows based on a names column, plus optional prediction distribution output.
• Model configuration bundles covering broad multi-national and targeted country-specific setups.
• PyTorch-based inference stack with configurable device, batch-size, and model selection.

Workflow

• Choose a model family depending on classification granularity and target population.
• Run batch or single-name inference with optional full probability distributions.
• Export prediction outputs to CSV for review, QA, and downstream analytic use.

Outputs

• Configurable ethnicity prediction pipeline with command-line execution modes.
• Comparative model results across several predefined label systems.
• Portable prediction outputs for bias analysis and exploratory empirical workflows.