Mining Web Data for Public Health

Опубликовано 21 июня 2016, 23:36

Recent years have seen the adoption of new Web data sources in a wide range of health areas. Of all areas, public health applications in behavioral medicine have the most potential to change how we conduct research, opening up exciting new opportunities. Fundamentally, behavioral medicine requires understanding how people make health decisions: what influences their decision, how they weigh information, and how social connections impact decisions. Web data sources provide new opportunities for studying these questions. Answering these questions often requires new data mining methods. In this talk, I will present multi-dimensional topic models of text which jointly capture topic and other aspects of text. We describe Factorial Latent Dirichlet Allocation, a multi-dimensional model in which a document is influenced by K different factors, and each word token depends on a K-dimensional vector of latent variables. I will demonstrate the advantages of this model in the application of mining drug experiences from web forums.