User:EpochFail/Grouplens Lab 2009-12-14

GroupLens Research is a research lab in the Department of Computer Science and Engineering at the University of Minnesota, Twin Cities specializing in recommender systems, online communities, mobile and ubiquitious technologies, digital libraries and local geographic information systems.

History
John Riedl and Paul Resnick originated the idea of GroupLens when they attended the Computer Supported Cooperative Work (CSCW) conference in 1992 and heard a keynote speaker Shumpei Kumon talking about an information economy. They felt that the types of technology that was proposed was too optimistic. John and Paul also felt that there was an abundance of information that was being wasted out there, especially the opinions that were generated by people about the information that they read on the internet. This idea is what led to the concept of GroupLens was born.

This lack of organized information fueled John Riedl and Paul Resnick to begin a system called GroupLens, which was a collaborative filtering system used for Usenet news that allowed for postings to be rated on a scale, then was distributed to partnering sites as a shadow newsgroup. These ratings allowed for the system to predict how much an individual would like an article, based on their ratings of previous articles. A feasibility test was done between MIT and the University of Minnesota and the published results were part of the CSCW conference of 1994.

In 1995, GroupLens Research expanded the team and hired Joseph Konstan and Bradley Miller and later others, to re-implement GroupLens over the internet with a centralized server, called the Better Bit Bureau. This version rolled out with an experimental group of around 200 users and the results showed that collaborative filtering worked and could predict better than overall averages. Collaborative filtering in this matter also changed the way people read articles - instead of merely skimming articles, they were reading them more closely to provide the most accurate rating.

In the Spring of 1996, the first workshop on collaborative filtering was put together by Paul Resnick (now at AT&T), and Hal Varian at the University of California, Berkeley. This workshop was financially supported by Infonautics. There, researchers from GroupLens and other projects that were studying similar systems came together to explore what each other were doing.

Later that Spring, a series of chance occurrences, commercial opportunities, and chance meetings lead to the founding of Net Perceptions by John Riedl, Bradley Miller, Joseph Konstan, Steve Snyder, and David Gardner). Net Perceptions went on to be one of the most successful recommender engine companies, but later became a casualty of the dot com crash.

Meanwhile, research continued at the University of Minnesota. The area of human–computer interaction became GroupLens, named after the first system of collaborative filtering by John Riedl and Paul Resnick. The department was given its first grant from the National Science Foundation to study algorithmic issues in collaborative filtering.

The Fall of 1997 saw a the launch of MovieLens. MovieLens uses "collaborative filtering" technology to make recommendations of movies/videos that you might enjoy, and to help you avoid the ones that you won't. Based on the ratings on the movies you have seen, MovieLens generates personalized predictions for movies you haven't seen yet. MovieLens, on the serious side, is a unique research vehicle for dozens of undergraduate and graduate students researching various aspects of personalization and filtering technologies. The idea of MovieLens came to fruition when DCC SRC's EachMovie site was closed down, and the researchers behind it gave out, for free, the anonymous data of two researchers who wanted to continue the project. Thanks to the speedy work of Brent Dahlen and John Herlocker, the two had the MovieLens system up and running in a months time. Since then, the site has had substantial success in attracting publicity and users. The site has been non-commercial since the beginning, and experimenters have been able to research collaborative filtering algorithms, interfaces, and other features.

Over the next dozen years, the GroupLens Research group continued to grow and diversify. Research areas have included algorithmic core of collaborative filtering, including item-based collaborative filtering and dimensionality reduction approaches, studies of new user experiences and how items can be selected to improve that experience, new applications of recommender systems and techniques. Most recent expansion dating back to 2002 has been the group expanding into social computing and online communities with the addition of Loren Terveen. GroupLens continued to take on new challenges in eliciting new contributions to online communities and shaping these communities to influence reputation. This work has involved a collaboration with the University of Minnesota, University of Michigan, and Carnegie Mellon University.

Currently
Current Research Grants:

This idea of organizing and rating
 * Net Perceptions

Movielens

 * MovieLens was launched in Fall of 1997.
 * The site has been non-commercial since the beginning, and experimenters have been able to research collaborative filtering algorithms, interfaces, and other features.
 * MovieLens uses "collaborative filtering" technology to make recommendations of movies/videos to users.
 * Based on the ratings on the movies you have seen MovieLens generates personalized predictions for movies you haven't seen yet.
 * DCC SRC's EachMovie site was closed down, and the researchers behind it gave out, for free, the anonymous data to researchers who wanted to continue the project.
 * From the provided data set, Brent Dahlen and John Herlocker built the MovieLens system in only a month.
 * Since then, the site has had substantial success in attracting publicity and users.
 * Note: Get exact amount at exact date referenced from some paper.

Wikipedia

 * Vandalism and value
 * Informal peer review mechanism
 * Quantitative analysis of Wikipedians and learning effects
 * Analysis of how policy affects what articles are deleted from Wikipedia + Wikipedia end-of-life predictions.