racebert 1.0.0

Race and Ethnicity Prediction from names

RaceBERT -- A transformer based model to predict race and ethnicty from names

Installation

Using a virtual environment is highly recommended!

pip install racebert

Usage

from racebert import RaceBERT

model = RaceBERT()

# Te predict race
model.predict_race("Barack Obama")

>>> {"label": "nh_black", "score": 0.5196923613548279}

# Predict ethnicity
model.predict_ethnicty("Arjun Gupta")

>>> {"label": "Asian,IndianSubContinent", "score": 0.9612812399864197}

GPU

If you have a GPU, you can speed up the computation by specifying the CUDA device when you instantiate the model.

from racebert import RaceBERT

model = RaceBERT(device=0)

# predict race in batch
model.predict_race(["Barack Obama", "George Bush"])

>>>
[
        {"label": "nh_black", "score": 0.5196923613548279},
        {"label": "nh_white", "score": 0.8365859389305115}
]

# predict ethnicity in batch
model.predict_ethnicity(["Barack Obama", "George Bush"])

HuggingFace

Alternatively, you can work with the transformers models hosted on the huggingface hub directly.

• Race Model: https://huggingface.co/pparasurama/raceBERT-race
• Ethnicity Model: https://huggingface.co/pparasurama/raceBERT-ethnicity

Please refer to the transformers documentation.