New paper: "In Your Own Words"! We propose a computational framework for identifying interpretable themes from free-text survey data, and demonstrate its benefits on a new dataset of self-described race, gender, and sexual orientation. 🧵1/