Welcome Guest | Sign-In | Register français

About Us

inDiscover (composed of the words Independent and Discover), is a tool whose main focus is helping independent musicians get heard by new audiences and to help people interested in music discover new and exciting music that they normally wouldn't hear in their everyday lives.

inDiscover uses collaborative filtering techniques and a rule engine to generate a list off recommended songs in the form of a playlist. By taking into account the way you have rated other songs, and how others have rated songs, we are able to predict how much you would like songs you have not rated. By applying rules to these predictions, we get a list of recommendations that we think you will like.

Below you will find a brief overview of how inDiscover can work for you:

For Everyone:
  • Once you register with us (a very quick process), you will be able have songs recommended to you based on your mood, location, and basic tastes in music.
  • By rating songs honestly in the multiple categories, we will be able to determine what you like and recommend you songs we think you will like and compose them into a playlist which you can download
  • The more songs you rate, the better we will be able to determine your tastes and our recommendation will become more accurate.
  • By getting people you know to join and rate some songs, this will also help increase the accuracy of our recommendations.
  • To get your recommendations in the various categories, use the PLAYLIST box on the right side of the page.


For Musicians:
  • inDiscover allows you to register as many artists as you would like.
  • Once you have registered an artist, you can then add songs under their account by providing a link to an mp3 file.
  • Users of inDiscover can then listen to, download, and rate your songs.
  • By rating your own songs and songs by other artists on inDiscover, you will increase the chances of your songs being recommended to other vistors. (Note: This does not mean to rate all your songs as highly as possible in every category! This will in fact skew the ratings and work against you! So try to be as reasonable as possible, it will help you in the long run.)
  • The more your song is rated by different users (and the more songs they have rated), the higher it will be weighted in the recommendation process and you will have a higher chance of being recommended to someone new.
  • So incourage your fans to come to the site, register, rate your songs, and then rate as many other songs as they can.


The Recommendation Process

Recommender systems have been around for a number of years. Web sites like Google and Amazon are fairly good at recommending web sites or books. In many ways, recommender systems have made the web the powerful tool it is today. However, these systems are essentially unidimensional: they try to guess overall what you need or want without taking into account very much the context. Context includes both the user context (what are you doing or want to do) but also other recommendations. We believe it is not enough to guess correctly what a user want, we need also to compose. Hence, inDiscover aims to provide high quality context-sensitive sets of recommendations based on explicit rating-based collaborative filtering. inDiscover is database-driven and leverages techniques from multidimensional databases (OLAP).

References:

Daniel Lemire and Anna Maclachlan, Slope One Predictors for Online Rating-Based Collaborative Filtering, SIAM Data Mining (SDM'05), 2005.
http://www.daniel-lemire.com/fr/abstracts/SDM2005.html

Michelle Anderson, Marcel Ball, Harold Boley, Stephen Greene, Nancy Howse, Daniel Lemire, Sean McGrath, RACOFI: Rule-Applying Collaborative Filtering Systems, Proceedings IEEE/WIC COLA'03, Halifax, Canada, October 2003. (NRC 46507)
http://www.ondelette.com/lemire/abstracts/COLA2003.html

Daniel Lemire, Scale and Translation Invariant Collaborative Filtering Systems. Information Retrieval, 7, pages 1--22, 2004. (NRC 46508)
http://www.ondelette.com/lemire/abstracts/IR2003.html

Daniel Lemire, Wavelet-Based Relative Prefix Sum Methods for Range Sum Queries in Data Cubes. Proceedings of CASCON 2002, Toronto, Canada, October 2002. (NRC 44967)
http://www.ondelette.com/lemire/abstracts/CASCON2002.html


RACOFI

inDiscover's technology comes from the RACOFI (Rule-Applying Collaborative Filtering) Composer system. RACOFI Composer is responsible for the algorithms that handle the rating and recommendation processes as well as any filtering rules which take place.
http://iit-iti.nrc-cnrc.gc.ca/projects-projets/racofi-composer_e.html