Date of presentation and deadline for the term paper: May 14, 2012
The last part of the course consists of a term paper, presentations and project 2 (which has a project report too).
1. The presentation is on two selected papers in recent conferences. You may combine the two papers in one presentation (30 min) or give two separate ones (40 min). The date for presentations for the whole class is May 14, Monday.
Presentation Topics:
Students |
Tentative Papers |
|
Jijie |
WWW’11, Information and Credibility on Twitter, |
WSDM’12: Correlating Financial Time Series by Microblogging Activity |
Billy |
RecSys’11: Crimewalker and Recommendation Models for Suspect Investigation |
RecSys’11: Yahoo Music Recommendations: Modeling Music ratings with Temporal Dynamics and Item Taxonomy |
Chi |
WWW’12: Using content and interactions for discovering communities in social networks |
WWW’12: Community detection in incomplete information networks |
Kan |
"OrdRec: An ordinal model for predicting personalized item rating distributions" Yehuda Koren |
? |
Fei |
Gideon Dror, Yehuda Koren, Yoelle Maarek, Idan Szpektor: I want to answer; who has a question?: Yahoo! answers recommender system. KDD 2011: 1109-1117 |
collaborative question and answering |
Raymond |
KDD’11 Sparsification of Influence Networks |
KDD’10 Scalable Influence Maximization for Prevalent Viral Marketing in Large-Scale Social Networks |
Simon |
Short and Tweet: Experiments on Recommending Content from Information Streams |
BALI Kanishk
2. Requirements for the Term Paper:
The term paper is a review of the selected papers (as opposed to the presentations, which present the overview of the papers). The term paper should not repeat the content of the paper, but give a summary and assessment. The length can be between 2 to 10 pages. The following points should be covered:
Overview of the paper: not to repeat, but give a high level summary of it.
Related works: did it miss any related works that are also important? Did any works come after it that are relevant?
Novelty: is it original? Does it use a technique that is outdated? Can you think of a better way?
Evaluation: are the metrics the paper chooses community standard? Are the data sets standard and complex/large-scale? Did it miss any baselines? Are there better methods that come after the paper is published, and you are curious on how they compare?