Difference between revisions of "2010 Challenge on Context-aware Movie Recommendation"

From RecSysWiki
Jump to navigation Jump to search
m (Reverted edits by JodieHyde (talk) to last revision by Alan)
 
(7 intermediate revisions by 3 users not shown)
Line 1: Line 1:
The '''2010 Challenge on Context-aware Movie Recommendation''' ('''CAMRA 2010''') was held at the [[RecSys2010|2010 Recommender Systems Conference]] and consisted of three context-specific tracks.
+
The '''2010 Challenge on Context-aware Movie Recommendation''' ('''CAMRA 2010''') was held at the [[RecSys2010|2010 Recommender Systems Conference]] and consisted of three [[context-specific]] tracks.
 +
 
 +
Participants were expected to submit papers, experimental results, and an additional alorithmic onepager describing their solution. The quality of all three was then used to crown the winner of the challenge. The overall winner was '''Mining Mood-specific Movie Similarity with Matrix Factorization for Context-aware Recommendation''' - [[Yue Shi]], [[Martha Larson]], [[Alan Hanjalic]].
  
 
== Tracks ==
 
== Tracks ==
 
=== Track 1 - Weekly Recommendation ===
 
=== Track 1 - Weekly Recommendation ===
The Weekly Recommendation track focused on recommending movies for the Christmas 2009 week and the Oscars' 2010 week in two datasets, both released exclusively for the challenge. One from [http://moviepilot.com Moviepilot] and one from [http://filmtipset.se Filmtipset].
+
The Weekly Recommendation track focused on [[recommending movies]] for the Christmas 2009 week and the Oscars' 2010 week in two datasets, both released exclusively for the challenge. One from [http://moviepilot.com Moviepilot] and one from [http://filmtipset.se Filmtipset].
 +
 
 +
Winning submission: '''Time vs. Tags: Factorization Models for Context-/Time-Aware Movie Recommendations''' - [[Zeno Gantner]], [[Steffen Rendle]], [[Lars Schmidt-Thieme]]
  
 
=== Track 2 - Mood-based Recommendation ===
 
=== Track 2 - Mood-based Recommendation ===
The mood-based recommendation track focused on recommending movies with a specific '''emotion''' keyword. The dataset used was an anonymized subset of the data collected by Moviepilot.
+
The [[mood-based recommendation]] track focused on recommending movies with a specific '''emotion''' keyword. The dataset used was an anonymized subset of the data collected by Moviepilot.
 +
 
 +
Winning submission: '''Mining Mood-specific Movie Similarity with Matrix Factorization for Context-aware Recommendation''' - [[Yue Shi]], [[Martha Larson]], [[Alan Hanjalic]]
  
 
=== Track 3 - Social Recommendation ===
 
=== Track 3 - Social Recommendation ===
The social recommendation track focused on recommending movies based on users' social connections, the dataset was an anonymized subset of the data collected by Filmtipset.
+
The [[social recommendation]] track focused on recommending movies based on users' social connections, the dataset was an anonymized subset of the data collected by Filmtipset.
 +
 
 +
Winning submission: '''Adapting Neighborhood and Matrix Factorization Models for Context Aware Recommendations''' - [[Nathan N. Liu]]
 +
 
 +
=== Live Evaluation ===
 +
A final evaluation was conducted in which the participants were asked to perform recommendations for a selected number of users for the context '''end of september''' (the week of the [[RecSys2011]] week). The first few movies in each recommended list were sent to users of [[Moviepilot]]'s service who had signed up for this. Each of these users were asked to answer the following questions<ref>[http://www.dai-labor.de/camra2010/live-evaluation-survey-questions/ Live evaluation survey questions]</ref>:
 +
* Was it a good recommendation?
 +
* Was it a surprising recommendation?
 +
* Did the movie match your expectations?
 +
* Do you think you would have watched the movie eventually, even if we hadn’t recommended it to you?
 +
 
 +
The winners of the live evaluation track was '''Time vs. Tags: Factorization Models for Context-/Time-Aware Movie Recommendations''' - [[Zeno Gantner]], [[Steffen Rendle]], [[Lars Schmidt-Thieme]].
 +
 
 +
== ACM TIST Special Issue ==
 +
Extended versions of some of the papers accepted to the workshop will be available in an ACM TIST Special issue<ref>[http://tist.acm.org/CFPs/TIST-SI-CAMRa.pdf CFP ACM TIST Special Issue on Context-Aware Movie Recommendation]</ref>.
  
 
== External links ==
 
== External links ==
 
* [http://www.dai-labor.de/camra2010/ Challenge Website]
 
* [http://www.dai-labor.de/camra2010/ Challenge Website]
 
* [http://portal.acm.org/citation.cfm?id=1869652&picked=prox&cfid=6592400&cftoken=81346418 Proceedings of the 2010 Challenge on Context-aware Movie Recommendation at the ACM Digital Library]
 
* [http://portal.acm.org/citation.cfm?id=1869652&picked=prox&cfid=6592400&cftoken=81346418 Proceedings of the 2010 Challenge on Context-aware Movie Recommendation at the ACM Digital Library]
 +
== References ==
 +
<references/>
  
 
{{DEFAULTSORT:CAMRA}}
 
{{DEFAULTSORT:CAMRA}}
 
[[Category: Competition]]
 
[[Category: Competition]]
 
[[Category: Movie recommendation]]
 
[[Category: Movie recommendation]]
 +
[[Category: Special issue]]
 +
[[Category: Event]]

Latest revision as of 12:24, 23 January 2012

The 2010 Challenge on Context-aware Movie Recommendation (CAMRA 2010) was held at the 2010 Recommender Systems Conference and consisted of three context-specific tracks.

Participants were expected to submit papers, experimental results, and an additional alorithmic onepager describing their solution. The quality of all three was then used to crown the winner of the challenge. The overall winner was Mining Mood-specific Movie Similarity with Matrix Factorization for Context-aware Recommendation - Yue Shi, Martha Larson, Alan Hanjalic.

Tracks

Track 1 - Weekly Recommendation

The Weekly Recommendation track focused on recommending movies for the Christmas 2009 week and the Oscars' 2010 week in two datasets, both released exclusively for the challenge. One from Moviepilot and one from Filmtipset.

Winning submission: Time vs. Tags: Factorization Models for Context-/Time-Aware Movie Recommendations - Zeno Gantner, Steffen Rendle, Lars Schmidt-Thieme

Track 2 - Mood-based Recommendation

The mood-based recommendation track focused on recommending movies with a specific emotion keyword. The dataset used was an anonymized subset of the data collected by Moviepilot.

Winning submission: Mining Mood-specific Movie Similarity with Matrix Factorization for Context-aware Recommendation - Yue Shi, Martha Larson, Alan Hanjalic

Track 3 - Social Recommendation

The social recommendation track focused on recommending movies based on users' social connections, the dataset was an anonymized subset of the data collected by Filmtipset.

Winning submission: Adapting Neighborhood and Matrix Factorization Models for Context Aware Recommendations - Nathan N. Liu

Live Evaluation

A final evaluation was conducted in which the participants were asked to perform recommendations for a selected number of users for the context end of september (the week of the RecSys2011 week). The first few movies in each recommended list were sent to users of Moviepilot's service who had signed up for this. Each of these users were asked to answer the following questions<ref>Live evaluation survey questions</ref>:

  • Was it a good recommendation?
  • Was it a surprising recommendation?
  • Did the movie match your expectations?
  • Do you think you would have watched the movie eventually, even if we hadn’t recommended it to you?

The winners of the live evaluation track was Time vs. Tags: Factorization Models for Context-/Time-Aware Movie Recommendations - Zeno Gantner, Steffen Rendle, Lars Schmidt-Thieme.

ACM TIST Special Issue

Extended versions of some of the papers accepted to the workshop will be available in an ACM TIST Special issue<ref>CFP ACM TIST Special Issue on Context-Aware Movie Recommendation</ref>.

External links

References

<references/>