Fall 2005 Reading Group
Objectives
The aim of this reading group is to keep ourselves abreast of the latest developments and trends in Information Retrieval, Data Mining, Machine Learning and other related areas of research. Towards this objective, we will aim at reading one paper each week from recent proceedings of SIGIR, CIKM, KDD, ICML, NIPS, AAAI, IJCAI and other related conferences and journals. In order to encourage breadth, paper discussion will consist of one section focusing on a single paper and an occassional second section quickly reviewing several potentially interesting papers.
This semester's reading group will be expanded to include short, informal student research talks. These talks are intended to keep peers aware of each other's work as well as providing a venue for students to get feedback in the early stages of their work.
Schedule
Tuesdays 4:00pm-5:00pm
CS243 (unless otherwise noted)
| Date | Presentor | Talk | Notes |
|---|
| 09.13.05 | ... | ... | organizational meeting |
| 09.20.05 | FernandoDiaz | Active Feedback | |
| | MarkSmucker | The Loquacious User: A Document-Independent Source of Terms for Query Expansion, Diane Kelly, Vijay Deepak Dollu, Xin Fu. SIGIR 2005. PDF | |
| 09.27.05 | MarkSmucker | Rocchio, IDF, and Language Modeling Retrieval (Research presentation) | CS151 |
| 10.04.05 | Hema Raghavan | Correlation of Feature Complexity and Active Learning performance (Research presentation) | CS151 |
| | DonMetzler | Boosting Web Retrieval through Query Operations, Mishne, Gilad and de Rijke, Maarten, ECIR 2005 PDF | CS151 |
| 10.11.05 | DonMetzler | Do We Really Need TF? (Research presentation) | |
| 10.18.05 | google research talk |
| 10.25.05 | RameshNallapati | A fast Approximation to Dirichlet distribution for modeling Text (Research Presentation) | |
| 11.01.05 | CIKM |
| 11.08.05 | | | |
| 11.15.05 | TREC |
| 11.22.05 | | | |
| 11.29.05 | | | |
| 12.06.05 | | | |
| 12.13.05 | | | |
| 12.20.05 | | | |
Format
Each meeting will consist of two parts: a research talk and a paper talk.
Research Talk
A student will be responsible for presenting a 15 minute research talk with 5 minutes reserved for questions. The intent of this talk is to keep students aware of your
current research. Please don't present material which is not significantly different from your previous work (this also is meant to dissuade the time from being used for conference practice talks).
Paper Discussion
A second student will be responsible for leading a 15-20 minute discussion of a new and (hopefully) interesting paper. We expect the group to have read the paper before the meeting so please don't spend 5 minutes reviewing what we've already read.
Ramesh proposed the following "democratic" algorithm for choosing papers to read which worked well in Summer 2005,
- Anyone is free to post papers to this web site at any time.
- Group members are expected to cast their votes to a paper that they find interesting by signing their name below the paper. Members are expected to check this page frequently and cast their votes whenever a new paper is seen.
- For any given week, the most popular paper by the end of the previous week will be chosen for reading. In case of a tie, the earliest posted paper among the tied ones will be chosen so as to encourage people to post as early as possible. Please insert your paper at the end of the list to maintain the time order of posting.
- The member who posted the paper selected for reading will be expected to lead the discussion on that paper, unless otherwise stated.
Papers and votes are carried over from Summer 2005.
Paper Reviews
At times, a third student may volunteer to reviewe several papers at a high level. The parameters are roughly 5 papers/10 minutes. The intent of this talk is to keep students aware of relevant research at other sites (eg, a sampling of good SIGIR papers, several interactive information retrieval papers). Papers presented here will probably end up on the potential paper list.
Proposed papers for future reading: Please post & vote below
- Learning to estimate query difficulty: including applications to missing content detection and distributed information retrieval, Elad Yom-Tov, Shai Fine, David Carmel, Adam Darlow PDF
- The maximum entropy method for analyzing retrieval measures, Javed A. Aslam, Emine Yilmaz, Virgiliu Pavlu, SIGIR 2005. PDF
- Text Classification with Kernels on the Multinomial Manifold, Dell Zhang, Xi Chen, Wee Sun Lee, SIGIR 2005. PDF
- Information Retrieval System Evaluation: Effort, Sensitivity, and Reliability, Mark Sanderson, Justin Zobel. SIGIR 2005. PDF
- Do Summaries Help?, Kathleen McKeown?; Rebecca Passonneau; David Elson; Ani Nenkova; Julia Hirschberg. SIGIR 2005. PDF
- Accurately Interpreting Clickthrough Data as Implicit Feedback, Thorsten Joachims, Laura Granka, Bing Pan, Helene Hembrooke, Geri Gay. SIGIR 2005. PDF
- Learning to extract information from semi-structured text using a discriminative context free grammar, Mukund Narasimhan, Paul Viola. SIGIR 2005. PDF
- Personalizing Search via Automated Analysis of Interests and Activites, Jaime Teevan, Susan T. Dumais and Eric Horvitz. SIGIR 2005. PDF
- A Conditional Random Field for Discriminatively-trained Finite-state String Edit Distance, Andrew McCallum?, Kedar Bellare and Fernando Pereira. UAI 2005. PDF
- Query-sensitive similarity measures for the calculation of interdocument relationships, Tombros,A. van Rijsbergen,C.J.. CIKM 2001.PDF
- A Study of Factors Affecting the Utility of Implicit Relevance Feedback, Ryen W. White, Ian Ruthven, and Joemon M. Jose. SIGIR 2005. PDF
- Multi-way Distributional Clustering via Pairwise Interactions, R. Bekkerman, R. El-Yaniv and A. McCallum?. In Proceedings of ICML 2005.PDF
- Query Expansion Using Random Walk Models, Kevyn Collins-Thompson and Jamie Callan, CIKM 2005 PDF
- Structured Queries in XML Retrieval, Jaap Kamps, Maarten Marx, Maarten de Rijke, Borkur Sigurbjornsson, CIKM 2005 PDF
- Relevance Weighting for Query Independent Evidence, Nick Craswell, Stephen Robertson, Hugo Zaragoza, and Michael Taylor, SIGIR 2005 PDF
- A Support Vector Method for Multivariate Performance Measures, Thorsten Joachims, ICML, 2005. PDF
to top