Stereotypical Gender Associations in Language Have Decreased Over Time

Current Version: 0.10

Release Date: September 28, 2018

Please contact moamin@cs.stonybrook.edu for quick response to resolve any bug or feature update.

Reproducing Results Discussed in Manuscript

Please follow the following steps to download source code:

  • Download and install git and git's extenstion git-lfs
  • Clone source code: git clone https://github.com/ruhulsbu/StereotypicalGenderAssociationsInLanguage.git
  • Go to the StereotypicalGenderAssociationsInLanguage directory
  • Unzip the dataset.tar.gz using command: tar -xvzf dataset.tar.gz
  • Run jupyter in current directory using command: jupyter notebook

Open the following notebook files:

  • Significance Test on Gender Bias using WEAT Topics.ipynb
  • Stereotypical Gender Association in Language Over Time.ipynb
  • Run each cell in the notebooks sequentially to reproduce the results

Dependencies

Large file dependencies:

  • git-lfs to download large file dataset.tar.gz

This dataset is originally hosted in Histwords Project. The dataset can be downloaded from this link: Embeddings. Please extract the files into the directory dataset in project home directory.

Core dependencies:

  • python 3.6.0
  • numpy
  • scipy
  • sklearn
  • seaborn
  • jupyter