metadata

language:
  - en
configs:
  - config_name: words
    data_files:
      - split: words
        path: words.parquet
  - config_name: associations
    data_files:
      - split: associations_weat
        path: associations_weat.parquet
  - config_name: associations_wefat
    data_files:
      - split: associations_wefat
        path: associations_wefat.parquet

Usage

When downloading, specify which files you want to download and set the split to train (required by datasets).

from datasets import load_dataset

words = load_dataset("fairnlp/weat", data_files=["words.parquet"], split="train")
associations = load_dataset("fairnlp/weat", data_files=["associations_weat.parquet"], split="train")

Dataset Card for Word Embedding Association Test (WEAT)

This dataset contains the source words of the original Word Embedding Association Test (WEAT) as described by Caliskan et. al. (2016).

Dataset Details

The dataset contains word lists and attribute lists used to compute several WEAT scores for different embedding associations. For details on the methodology, please refer to the original paper. This dataset is contributed to Hugging Face as part of the WEAT implementation in the FairNLP fairscore library.

Dataset Sources

Paper [optional]: lcs.bath.ac.uk/~jjb/ftp/CaliskanSemantics-Arxiv.pdf

BibTeX:

@article{DBLP:journals/corr/IslamBN16,
  author       = {Aylin Caliskan Islam and
                  Joanna J. Bryson and
                  Arvind Narayanan},
  title        = {Semantics derived automatically from language corpora necessarily
                  contain human biases},
  journal      = {CoRR},
  volume       = {abs/1608.07187},
  year         = {2016},
  url          = {http://arxiv.org/abs/1608.07187},
  eprinttype    = {arXiv},
  eprint       = {1608.07187},
  timestamp    = {Sat, 23 Jan 2021 01:20:12 +0100},
  biburl       = {https://dblp.org/rec/journals/corr/IslamBN16.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}