This folder contains code to generate the HolisticBias dataset, a set of sentences containing demographic
identity language (e.g. “Hi! I am a Catholic grandmother.”), used in the context of a two-person conversation.
Sentences are formed by combining (1) an identity term from one of 13 demographic axes, (2) a noun referring to
a person (mom, boy, grandparent, etc.), and (3) one of several dozen sentence templates.
