WEAT-WEFAT

Repo for research tasks associated with WEAT association tests, low resource word embeddings, etc The associated research papers are available here:

• Semantics derived automatically from language corpora contain human-like biases
• Detecting Emergent Intersectional Biases: Contextualized Word Embeddings Contain a Distribution of Human-like Biases
• ValNorm: A New Word Embedding Intrinsic Evaluation Method Reveals Valence Bases are Consistent Across Languages and Over Decades

Contents

1. lib/weat.py

This python file contains the Word Embedding Association Test (WETA) and Work Embedding Factual Association Test (WEFTA) implementations.

2. main.ipynb

This jupyter notebook contains the main file that runs the WETA on 10 datasets that were used to conduct Implicit Association Tests (IAT).

3. targets_attributes_data/

This folder contains the 10 datasets that each have 2 sets of targets and attributes.

Future Implementation

1. The code can be improved to add tests for WEFTA.