MusicStrands Uses Artificial Intelligence For Music Recommendations

MusicStrands is a recommendation system which recommends user music based on artificial inteligence.Users of the MusicStrands website have access to a directory of 3.7 million songs. They can search the collection and receive music recommendations according to their own tastes in music.The artificial intelligence techniques used for the recommender systems are based on statistical learning, Bayesian estimation, probabilistic reasoning and visualisation techniques. Five patents have already been awarded thanks to the innovations, and it is hoped that another forty will also be granted over the next five years. For more details goto

http://www.sciencedaily.com/releases/2005/02/050224121427.htm

Good Ratings Gone Bad: Study Shows Recommender Systems Can Manipulate Users' Opinions

This is a news article published in "Science Daily" on 8 april 2003. It says that according to a study conducted by University of Minnesota researchers a system which lies about user ratings can manipulate user's opnions. They conducted three different experiments where they tried different ratings and predictions for movies and examined how other users opnion changes according to this changes. For full article Click Here.

series amazon recommendation

it is a special feature that recommends based on the kind of you interact with the system. If you just watch the products it recommends the similer items. If you add it to wishlist it will show items which are in series it terms of use.

User XQuery Pattern Method based Personalization Recommender Service

This paper presents a semantic web recommender system for personalization service using a user xquery method. The user XQuery provides an important recommendation service that assists ecommerce and m-commerce by creating personalized document retrieval and recommender systems. In order to provide these services effectively, the user XQuery
should be a "user orientated query pattern". This is achieved through the utilization of a XML based e-commerce and mcommerce using the Semantic Web, including XML Query patterns, mining XQuery streams, personalization, user profiling and other XML based technologies.

For full detais visit IEEE library

IMPl. OF KNOWLEDGE-BASED RECO. SYS using JESS

This is a resource which initially describes the difference between collaborative filtering and knowledge based recommender system. The it explains how java expert shell system implement this notion using knowledge base of the product domain, inference engine and user interface module.

The detail document can be found at IEEE Library

Asynchronous Recommendation System

Current recommender systems have matches the taste of the users or items with other item and make decision about recommandation. But there are certain problems regarding commitee of users, seperability of assumptions and syncrony. Asynchronous recommendation system solves these problems. Visit ACM library for full details.

Utility-Based Neighbourhood Formation

This paper proposes novel neighbourhood formationand similarity weight transformation schemes for automatedcollaborative filtering systems. it demonstrate the benefitsof new schemes from the point-of-view of the efficiency androbustness provided, while achieving the accuracy and coverageof a benchmark k–Nearest Neighbour (k–NN) model.

Source: ACM Library

Amazon.com Recommendations-item to item collaborative filtering

Amazon.com used item to item collaborative filltering which has online computation scale independantly of number of users and number of items in catalog. Unlike traditional collaborative filtering where system finds similer set of customer who has purchased same items as current customer, amazon.com tries to find items which are similer to one that current customer has already purchased,viewed or recommended.

The entire report can be found here.

http://www.win.tue.nl/~laroyo/2L340/resources/Amazon-Recommendations.pdf

SVD Recommendation System in Ruby

SVD recommendation system was used to rank the user feedback gor the Big Family Fuys show.

The Entire News Can be found here.
http://www.igvita.com/2007/01/15/svd-recommendation-system-in-ruby/

PHOAKS: a system for sharing recommendations

PHOAKS (People Helping One Another Know Stuff) is an experimental system designed to help
users locate information on the World Wide Web. Using a collaborative filtering approach, it sifts
through Usenet news messages and automatically identifies, keeps track of and redeploys
recommendations of Web resources. PHOAKS differs from other recommender systems in that it
is based on the principles of role specialization and reuse.

Here is the link for the document: http://www.ischool.utexas.edu/~i385d/readings/Terveen_PHOAKS_97.pdf

An MDP-Based Recommender System

Hi

I found a new recommender system called "Markov Decision Process Based Recommendation System". Unlike typical recommender system which focuses on static view of recommendation system, MDP based recommender system takes into account the long term effect of each recommendation and the expected value of each recommendation. The entire document can be found at

http://jmlr.csail.mit.edu/papers/volume6/shani05a/shani05a.pdf

Thanks

Privacy Risks in Recommender Systems

There is always a dillema about where the boundry of personalization ends and when it breaks someone's privacy. The following paper uses a graph-theoretic model to study the benefit from and risk to straddlers.

http://people.cs.vt.edu/~naren/papers/ppp.pdf