Biasly
Artificial intelligence (AI) to detect and correct misogynistic language.
About the Project
Biasly is an AI research project leveraging cutting edge natural language processing algorithms to identify and correct misogynistic language in written text, whether such text is expressed overtly or covertly.
Biasly is being developed to explain why language can be interpreted as misogynistic and suggest ways of rephrasing to reduce or remove the misogynistic implications.
How it Started
Biasly was originally devised by Andrea Jang, Carolyne Pelletier, Ines Moreno and Yasmeen Hitti during their summer training at the AI4Good Lab. Their goal was to build a machine learning tool capable of detecting and correcting gender bias in written text. To bring their idea to fruition, the team decided to undertake a Humanitarian AI Internship with Mila, where they developed a taxonomy for gender bias in text. Although their internship has since ended, Mila’s team of researchers has continued to pursue this work to fulfill Biasly’s vision.
Taking Biasly to a New Level
Biasly’s current team has been strategically assembled to ensure the responsible and high impact delivery of this work.
Our ML researchers are Anna Richter* (MSc in Cognitive Science) and Brooklyn Sheppard* (MSc in Speech and Language Processing).
The project is being managed by Senior Projects Manager, Allison Cohen, in collaboration with experts from the field; including Linguistics expert, Dr. Elizabeth Smith; Gender Studies specialist, Dr. Tamara Kneese; Natural Language Processing Advisor, Dr. Yue Dong and Language Processing Engineer, Carolyne Pelletier.
The project has also been fortunate to have had the insight and expertise of AI Researcher, Dr. Ioana Baldini and Gender Studies Expert, Dr. Alex Ketchum.
Dataset
In 2024, the Biasly team released the complete dataset of misogynistic sentences. You can access it using the link below:
Resources
BiaSWE: An Expert Annotated Dataset for Misogyny Detection in Swedish
Inspired by the Biasly project, AI Sweden has created a misogynistic dataset for the Swedish language.
Learn more
https://arxiv.org/html/2502.07637v1
Mila’s AI4Humanity and Biasly: Best Practices to Develop Socially Responsible AI
A case study on Biasly by the Rotman School of Business (University of Toronto)
Read the case study
https://mailchi.mp/112a7c42d338/tgduedfz9g
Subtle Misogyny Detection and Mitigation: An Expert-Annotated Dataset
Publication from the project presented at the ACL conference.
Read the article
https://arxiv.org/abs/2311.09443
Subtle Misogyny Detection and Mitigation: An Expert-Annotated Dataset
Publication from the project presented as a Spotlight at NeurIPS.
Read the article
https://openreview.net/forum?id=5hY19x1hdq
Meet the Team
Mila Members
Collaborators
Dr. Yue Dong (University of California, Riverside)
Dr. Tamara Kneese (University of California, Berkeley)
Carolyne Pelletier (Mantium)
Dr. Elizabeth Smith (UQÀM)
*Dr. Ioana Baldini (former collaborator, IBM Research)
*Dr. Alexandra Ketchum (former collaborator, McGill University)
*Anna Richter (former collaborator, Mila)
*Brooklyn Sheppard (former collaborator, Mila)
News
Partners
This project has benefited from the generous support of DRW.
