This event has ended. View the official site or create your own event → Check it out
This event has ended. Create your own
View analytic
Friday, October 14 • 1:10pm - 1:50pm
Combining Content and Collaboration in Recommenders
Recommender Systems are typically built on two different types of training data: historical user-engagement, and the textual content of the items themselves (either descriptive text, tags, structured metadata, or the actual raw content of text items on their own). This talk is an introductory overview of how to build a recommender system which uses both types of inputs to build a “mixed-mode” recommender, where you can parameterize (at request time, in some cases!) how much you want to rely on content, and how much on collaborative filtering. We’ll walk through building a horizontally scalable parameterized recommender service from just three components: Solr, Spark, and of course: training data.

avatar for Jake Mannix

Jake Mannix

Lead Data Engineer, Lucidworks
Living at the intersection of IR and applied ML, Jake likes to build “data-driven products” like personalized search engines, recommender systems, and relevance-enhancing substrate like user-interest classifiers. Currently the lead data engineer at Lucidworks, doing relevance R&D in the office of the CTO, Jake previously was a tech lead and founding member of the user interest modeling and account search teams at Twitter, and... Read More →

Friday October 14, 2016 1:10pm - 1:50pm
Commonwealth Sheraton Boston