LoginSign Up
← Back to all sparks
G

genderBR

INFRA · APIS
Velocity0.0

Predict Gender from Brazilian First Names

github.com/meirelesff/genderbr ↗

A census lookup table learns to guess names it has never seen.

demographicsneural-inferencebrazilcensus-datatorch
◆Current state
genderBR infers gender from Brazilian first names, and until this year it did so purely by looking names up in IBGE census frequency data. Version 1.3.0 adds a second, learned path: a character-level neural network that scores names the census never recorded. The package now carries torch as a hard dependency and pulls model weights from Hugging Face on first use.
◆Where it's heading
The arc is from data lookup to inference. 1.2.0 modernised the lookup side by adding 2022 census data and swapping the dplyr join layer for data.table; 1.3.0 keeps that intact and bolts a model beside it rather than replacing it. The maintainer is also cleaning up platform-dependent string handling and deprecating the encoding argument, which suggests consolidation around the new code path.
◆Prediction
Expect the deprecated encoding argument to be removed and the neural path to gain the threshold-tuning controls the census path already has. Whether get_gender_nn() becomes the default is the open question the release notes do not answer.

◆Recent moves

  1. 4mo ago

    Neural name model predicts gender beyond the census list

    ⚡ SPARK

    The release that changes what kind of package genderBR is: alongside the census lookup it now ships get_gender_nn(), a character-level neural network that generalises to names absent from IBGE data, with cache management and Hugging Face weight downloading in support.

    View source ↗
  2. 7mo ago

    2022 census data lands, data.table replaces the dplyr joins

    The lookup path gets a decade newer: a year argument exposes 2022 IBGE names data at national and state level, and the internal dataset now carries both 2010 and 2022 probabilities for offline use. The dependency trail shrinks as magrittr goes and data.table takes over the joins.

    View source ↗
  3. 5y ago

    Adds unit and input tests with minor bug fixes

    A maintenance patch adding internal input tests and unit tests with a few small bug fixes. Nothing user-facing changes; it is the hardening step before the data and model work of the following releases.

    View source ↗