Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

RecSys 2014

No description
by

Michele Trevisiol

on 5 January 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of RecSys 2014

cold-start
.
Cold-start News Recommendation with Domain-dependent BrowseGraph
Michele Trevisiol
, L.M. Aiello, R. Schifanella, A. Jaimes

RecSys 2014
BrowseGraph and Referrer URL
.
good
collective indicator
of content
interest
.
nodes are
pages
edges are aggregated
browsing transitions
external layers
add useful information
user interest
Referrer URL
.
referrer
Yahoo News Browse Graph
.
2 months of Yahoo News logs
~500M pageviews
referrer
URL
YNews BrowseGraph
Users' browsing session
from sessions to BrowseGraph
.
Social Network
Search Engine
Kendall-tau of Nodes PageRank
session length
structure of the graphs
nodes overlap
articles ranking
topics of the news articles
very different size
well connected
Jaccard similarity of node sets
Sessions Length
.
short sessions as expected
Referrer Graphs Structure
.
Nodes Overlap and Importance
.
Topics
.
7
.
8
.
9
.
10
.
Study the Referrer Graphs
.
Referrer Graphs
.
Recommender Task
.
Experiment Setup
.
techniques
[rnd] Random
[pop] Popular
[edge] Edge
[cb] Content-based
graph selection
[ref] Referrer Graph
[full] Browse Graph
[mix] ref > full
hourly time slots (2 months)
previous hour : build the model
current hour : test set
dataset
Browse Graph :
very useful as
indicator of implicit user feedback

Referrer Graphs :
show a
good potential
of selecting good candidates
Recommending the next news article
.
Evaluation : Precision@1
.
Evaluation : MRR@3
.
14
.
15
.
Conclusion and Future Work
.
user cold-start
.
common problem that never goes away
general solution : Popularity? CB? Freshness?
user browsing traces
waiting for some user traffic?
(Park and Chu, RecSys 2009; Konstan and Riedl, 2012)
referrer
URL
BrowseGraph
User browsing session
news domain
.
huge traffic:
major portion
of the web traffic
thousands of new articles per day
short user sessions
(The Guardian, 2012)
(Kumar and Tomkins, SIGIR 2010)
(Tsagkias and Blanco, SIGIR 2012)
referrer URL
can we face the cold-start problem exploiting this information?
(full) BrowseGraph
full
graph
This approach could be applied on
different domains
and
type of items
characterization of the user session by the referrer URL
(Chiarandini et al., ICME 2012)
extensions :
add
extra implicit
information
(IP, cookies, etc.)
integration with
user profile

(does it change by referrer?)
additional methods of node selection
BrowseGraph
Full transcript