Bias metrics for gender based on arXiv:1910.10872 paper

[mariojose123]

Is your feature request related to a problem? Please describe.

I want to create a library to get the metrics of arXiv:1910.10872 given a model o names entity recognition, but i think it would be good if its implemented on fairlearn instead of other library.

Describe the solution you'd like

Given a model of name entity recognition, templates according to the paper or created by the user and frequency of name in a region, I want to calculate all errors described in this paper for gender bias analysis,

Describe alternatives you've considered, if relevant

Additional context

example of a plot of the gender bias in many templates given the frequency of a names in Brazil according to IBGE(a Brazilian government association for Brazilian Statistics)

[mariojose123]

I already created a class for this type of metrics, but i don't know if this specific metric for name entity recognition bias is outside of the scope of this library

[adrinjalali]

Thanks for opening this issue @mariojose123 . It's indeed a good question weather it's too specific or not. I'm leaning towards being a +1 on having a general version of this in the library. But we'll need to discuss with other @fairlearn/fairlearn-maintainers