People's tastes are their systems of aesthetic judgment and preference. Taste systems are generative; understanding the tastes of friends makes it possible to anticipate what book or movie they would like, or how they might feel about a particular news topic. We are pioneering a computational approach to automatically modeling the tastes of individuals and whole communities by mining their texts; blogs, social network profiles, and homepages; and analyzing the attitudes and preferences contained therein. By also applying machine-learning techniques, we generalize these models so that they can predict how a person would feel about hypotheticals, and even forecast tomorrow's trends.