Inspired by recent work from the University of Michigan, we built a knowledge engine for extracting disease severity from symptoms using speech input and models trained on wikipedia data.

The team who worked on PRIORI was able to build a pipeline for accessing severity of manic episodes by using speech data from bipolar patients; we sought out to build a more generalized model by leveraging TFIDF Vectors in a bayesian setting. This model leveraged CMU's NELL Knowledge Base and mined wikipedia for training text related to diseases and everyday life.

Share this project:

Updates