[Gala-l] tackling gender bias in text

[Inès Moreno]

Hello everyone,

I'm one of four research interns taking part in the humanitarian AI internship at the Montreal Institute for Learning Algorithms (MILA<https://mila.quebec/en/>). My team is working on a project to tackle gender bias in text with machine learning (a branch of artificial intelligence). In recent years, a lot of AI models are found to be biased because machines are trained with data that contains human-like biases.

To solve this problem, we are attempting to build a clean dataset that contains text labeled as 'gender biased' or 'not gender biased'. Our first prototype was launched at the end of the AI4GOOD summer lab and has been active ever since however, the main issue we had was context and people not fully understanding what gender bias is.

We really want to put emphasis on the definition and this is why we are reaching out to several individuals in the community. In order to build a good dataset, we need to understand what gender bias is and define it as precisely as possible.

 We are aware that, it is very complicated to give a dictionary-like definition of gender bias; for now, we have narrowed it down to 3 subcategories (stereotypes, gender generalizations, and abusive language) but we would like to confirm this with professionals.

Here is a questionnaire we have prepared which would help us affirm and clarify our understanding of gender bias https://goo.gl/forms/0qMohbYnMbUFp2ii1
I would really appreciate if you could take the time to answer it and if you are still unclear on what our purpose is please feel free to message me.

If you’d like more information on our work, we would be happy to send you our project proposal.

Thank you very much for your time,
Best,
Ines

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/gala-l/attachments/20181106/fe607f95/attachment.htm>