Activate 2018 has ended
Back To Schedule
Thursday, October 18 • 1:15pm - 1:55pm
Deep Learning for Unified Personalized Search and Recommendations

Log in to save this to your schedule, view media, leave feedback and see who's attending!

One of the really nice things about modernl neural network architectures is that they are easily capable of incorporating many different heterogenous sources as inputs. In this talk, we'll go over how to create a ranking model, trained on user-identified clicks, which can learn a pointwise ranking function from (user_id, query_string, document snippet) tuples, which takes into account the users' past query/click history to personalize the ranking to their preferences.

The modeling will be described in Keras (with python), and the runtime examples will use Tensorflow's Java API to allow easier integration with a Solr LTR plugin.

avatar for Jake Mannix

Jake Mannix

Chief Data Engineer, Lucidworks
Jake Mannix is the Chief Data Engineer at Lucidworks. Before joining Lucidworks, Jake worked on the Semantic Scholar project at the Allen Institute for Artificial Intelligence, and prior to that was tech lead for Twitter’s data science and data engineering teams, building both the... Read More →

Thursday October 18, 2018 1:15pm - 1:55pm EDT
Salon 4&5