I introduce an adapted embedding approximation technique that allows researchers to interpret how localized speech acts contribute to biases in global embedding spaces.
I thank Professor James Evans for his support in purchasing and sharing the COHA dataset.