Each paper presentation is allocated 10 minutes for the presentation, followed by 7 minutes for Q&A.

09:00-10:30    Session 1

09:00-09:05    Opening
09:05-09:22    Recommending Running Routes: Framework and Demonstrator. Benedikt Loepp et al.
09:22-09:39    Recommendations for sports games to bet on. Toon De Pessemier et al.
09:39-09:56    Time-aware Personalized Popularity in top-N Recommendation. Vito Walter Anelli et al.
09:56-10:13    Retrieving and Recommending for the Classroom: Stakeholders, Objectives, Resources, and Users. Michael Ekstrand et al.
10:13-10:30    User and Context Aware Composite Item Recommendation. Krista Drushku et al.

10:30-11:00    Coffee break
11:00-12:30    Session 2

11:00-11:17    Recommendations for repeat consumption products considering users’ tendency to repeat. Qian Zhang et al.
11:17-11:34    Finding Your Home: A Unique Recommendation Problem. Eric Ringger et al.
11:35-12:25    Music Recommendation System Unplugged. Shared KARS/ComplexRec keynote by Tao Ye (Pandora) and Vito Ostuni (Pandora)
12:25-12:30    Closing

12:30-                Lunch



Title: Music Recommendation System Unplugged.

Abstract: In today’s world, a recommendation system is an integral and expected component in any online service. This is especially true in a large scale entertainment streaming service such as Pandora, where the listeners have grown accustomed to more and more personalized music experience not only fitting their expressed interest, but also their situations, through recommendations. In constructing these recommendation algorithms and systems, practitioners often need to consider a complex array of constraints such as different play formats (station, playlist, thumbs, personalized playlist, a visual browse), query originations (typed search in mobile, voice command), user tiers (radio only, premium on-demand) and a variety of clients. Our lives as practitioners are made simultaneously more interesting and more hectic by these constraints.

We use three main scenarios to illustrate the complexities faced by the recommendation system at Pandora, and show how we make tradeoffs with freshness, diversity, listener taste relevance, contextual relevance, among many other requirements. Firstly, we share genome tagging, keyword tagging, and automatic genre tagging, the most important knowledge building blocks of recommendations and explanations. Secondly, we present a contextual multi-armed bandit algorithm for ranking different types of items, such as albums, curated music stations, playlists, and artists, grouped into modules on one scrollable page. Thirdly, we highlight personalized new release (The Drop) playlist generation and some of its challenges such as cold start, artist disambiguation and how we balance artist familiarity and discovery. Finally, we give a brief discussion of new challenges brought on by recently popular voice platforms and our initial approaches.

Since 2005, Pandora’s music recommendation platform has powered the well know machine learning driven personalized radio product and recent premium on demand product, complete with support to a host of clients from web, mobile, and voice platforms for 70+ million listeners in the U.S.

Dr. Tao Ye is a Principal Scientist and Sr. Manager of Science at Pandora. She is a founding member of the Pandora science team, and has been working on personalized recommendation systems, measurements, and user modeling since 2010. Most recently, she has been leading the personalization and discovery science team that advances the machine learning and data driven innovations in search, voice interface, and many personalized music recommendation features in Pandora. She has two decades of experience in the software industry, holding research scientist and lead engineer positions in social media, networking and mobile systems. She holds 14 granted patents and has published 12 peer reviewed papers.
She received her PhD from University of Melbourne in Electrical and Electronic Engineering, her MS from UC Berkeley in EECS and dual BS degrees from Stony Brook University in CS and Engineering Chemistry.

Dr. Vito Ostuni is Staff Scientist on the Pandora listeners science team working on search, voice and recommendation systems. He holds a Ph.D. in Electrical and Information Engineering from Polytechnic University of Bari, Italy. His main interests are recommender systems, search and applied machine learning. Vito is author of 15 peer reviewed papers and the recipient of a best paper award at the I-Semantics 2012 conference.


Accepted papers

  • Benedikt Loepp and Jürgen Ziegler. Recommending Running Routes: Framework and Demonstrator
  • Toon De Pessemier, Bram De Deyn, Kris Vanhecke and Luc Martens. Recommendations for sports games to bet on
  • Vito Walter Anelli, Joseph Trotta, Tommaso Di Noia, Eugenio Di Sciascio and Azzurra Ragone. Time-aware Personalized Popularity in top-N Recommendation
  • Michael Ekstrand, Ion Madrazo Azpiazu, Katherine Landau Wright and Maria Soledad Pera. Retrieving and Recommending for the Classroom: Stakeholders, Objectives, Resources, and Users
  • Krista Drushku, Alexandre Chanson, Ben Crulis, Nicolas Labroche and Patrick Marcel. User and Context Aware Composite Item Recommendation
  • Qian Zhang, Koki Nagatani, Masahiro Sato, Takashi Sonoda and Tomoko Ohkuma. Recommendations for repeat consumption products considering users’ tendency to repeat
  • Eric Ringger, Alex Chang, David Fagnan, Shruti Kamath, Ondrej Linda, Wei Liu, Imri Sofer, Nicholas Stevens and Taleb Zeghmi. Finding Your Home: A Unique Recommendation Problem

The proceedings of the ComplexRec 2018 workshop are available here.