Rephil: Text to Concepts with Machine Learning

When: Friday, November 4, 2016
Where: PGH 232
Time: 11:00 AM

Speaker: Mr. Russell Howes, Software Engineer, Google

Host: Dr. Carlos Ordonez

Rephil is a machine learning system used throughout Google to identify the concepts or topics that underlie a given piece of text. Rephil determines, for example, that "apple pie" is related to some of the same topics as "chocolate cake", but has little in common with "apple iphone". The concepts used by Rephil are not pre-specified, but are derived by an unsupervised learning algorithm running on massive amounts of text. This process results in a Rephil model--a giant Bayesian network with concepts as nodes. This talk will discuss the structure of Rephil models and the distributed machine learning algorithm that builds them from terabytes of data.

Bio:

Russell Howes has been a Software Engineer at Google since 2014. He works at Google in the Los Angeles office where he works with the Rephil team on projects to extract meaning from text. Before working at Google Russell majored in math at BYU, did graduate work in Applied Math at UCLA, and worked at a few other companies in the LA area. In his spare time he enjoys beach volleyball and home improvement projects.