Exploiting content, structure, and communities in recommender systems

Speaker:        Dr. Julian McAuley
                Stanford University

Title:          "Exploiting content, structure, and communities in
                 recommender systems"

Date:           Wednesday, 30 October 2013

Time:           3:00pm – 4:00pm

Venue:          Room 1504 (near lifts 25/26), HKUST

Abstract:

Recommender systems have transformed the way users discover and evaluate
products on the web. In order to recommend products (or webpages, ads,
etc.) to users we must uncover the implicit tastes of each user as well as
the properties of each product. In addition to users' ratings, there are
many sources of structure and side-information that may help with this
task. For example, how can the text of users' reviews be exploited to
better understand why users rated products the way they did? How can we
model users' expertise in settings where users have 'acquired tastes'? How
does the title of a product, or the community in which it is marketed,
influence its chance of being adopted? In this talk I'll present a series
of papers that try to answer such questions. To do so, we'll introduce
large-scale datasets and models from traditional review communities (such
as Yelp and Amazon), to niche communities (such as RateBeer and
CellarTracker), and social media communities (such as Reddit).


*******************
Biography:

Julian McAuley is currently a postdoctoral scholar at Stanford University,
where he works with Jure Leskovec on modeling the structure and dynamics
of social networks. His current work on recommender systems is concerned
with modeling user behaviour in online communities, especially in terms to
their linguistic and temporal dimensions. Previously, Julian received his
PhD from the ANU under Tiberio Caetano, with whom he worked on inference
and learning in structured output spaces. His work has recently been
featured in Time, Forbes, New Scientist, and Wired, among others.