Speaker: Blake Hunter (Claremont McKenna College)

Title: Topic Point Processes

Abstract:
The widespread use of social media has created a need for automated extraction of useful information marking detection of trends. This is essential to gain knowledge from the vastly diverse massive volume of data. This talk explores topic modeling and self-exciting time series models applied to social media microblog data. We use topic modeling and compressed sensing to discover prevalent topics and latent thematic word associations within topics.  Modeling tweet topics over time as temporal or spatial-temporal Hawkes process allows identification of self-exciting topics and reveals relationships between topics. This talk looks at applications to Twitter microblogs, text mining, image processing, and content based search.