GenderPerformr

Intro

GenderPerformr is the model release from the paper It’s going to be okay: Measuring Access to Support in Online Communities by Zijian Wang and David Jurgens (in proceedings of EMNLP 2018).

It is the current state-of-the-art method that predicts gender from usernames based on a LSTM model built in PyTorch (as of Sept. 2018).

See the project website for full details, including contact information.

Install

Use pip

If pip is installed, genderperformr could be installed directly from it:

pip install genderperformr

From raw

git clone the project and do:

python setup.py install

Dependencies

python>=3.6.0
torch>=0.4.1
numpy
unidecode

Usage and Example

predict

predict is the core method of this package, which takes a single username of a list of usernames, and returns a tuple of raw probabilities in [0,1] (0 - Male, 1 - Female), and labels (M - Male, N - Neutral, F - Female, empty string - others).

Simplest usage

You may directly import genderperformr and use the default predict method, e.g.:

>>> import genderperformr
>>> genderperformr.predict("AdamMcAdamson")
(0.019139649, 'M')

Construct from class

Alternatively, you may also construct the object from class, where you could customize the model path and device:

>>> from genderperformr import GenderPerformr
>>> gp = GenderPerformr()

# Predict a single username
>>> gp.predict("John")
(0.087956183, 'M')

# Predict a list of names
>>> probs, labels = gp.predict(["BarryCA67", "pizzamagic", "KatieZ22"])
>>> f"Raw probabilities are {probs}"
Raw probabilities are [0.03398224 0.5439474 0.93964571]
>>> f"Labels are {labels}"
Labels are ['M', 'N', 'F']

More detail on how to construct the object is available in docstrings.

Model using new data partition

If you want to use the model described in Supplemental Material using the new data partition, you may construct the object via

>>> gp = GenderPerformr(is_new_model=True)

All other usages remain the same.

Citation

@inproceedings{wang2018its,
       title={It's going to be okay: Measuring Access to Support in Online Communities},
       author={Wang, Zijian and Jurgens, David},
       booktitle={Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)},
       year={2018}
}

Contact

Zijian Wang (zij<last_name>@stanford.edu)

David Jurgens (<last_name>@umich.edu)