• A Brief Overview and Discussion on Online Music Discovery and Recommender Systems

    by  • November 1, 2012 • News, Reading Group

    This session serves as a forum for discussing various music discovery systems currently available on the world wide web. Examples of such systems are not restricted to only those that rely on audio features extracted from music data for making recommendations, but also include those that rely on user & genre tags, data mining from music news websites, music blogs and data of various modalities. The session aims at creating an awareness of one context in which a considerable portion of research in Music Information Retrieval and Machine Learning for Music has been applied.

    The list below is only a small portion of such systems that are available on the world wide web, and not all details about all of them are known.

    It is the hope that in the course of, and following today’s meeting, this list can be made more complete and correct in its content, and updated with other similar services that others have come across or themselves used in the past. Group members are encouraged to come up and share any website not listed here during the meeting. Suggestions on how this list can be made more informative, or better organized are welcome.

    The following is an initial list that has been compiled to initiate the discussion:

    Service URL Discovery Demo
    Tag/Feature based


    Last.fm http://www.last.fm/ Recommendations are calculated using a
    Collaborative filtering algorithm that relies
    on user tags, MP3 tags uploaded via
    Media players using the Audioscrobbler
    Plugin. 

    Also allows manual recommendations by
    users to other users.
    We Are Hunted http://wearehunted.com/a/#/emerging/ Proprietary software for music recommendation that
    enables data aggregation, machine learning and
    content recommendation services. 

    Generates the music charts, powers the dynamic
    playlists, and delivers detailed analytics and reporting.

    Indexes approximately 100,000 music articles and
    40 million music related tweets each month.

    Actively monitors YouTube and the Facebook
    graph for artist activity.
    Musicovery http://musicovery.com/ Proprietary technology called “mood pad”
    that enables to position any song on a
    2 dimension continuous space. 

    Songs are described with 40 musical
    parameters that are based on music
    description and human acoustic perception.

    		 http://www.youtube.com/watch?v=zqkFUUWGH5E
    Quirk.fm http://www.quirk.fm/preview Idea to completely do away with genres. 

    Songs represented by a set of emotion-related
    descriptors.

    Do the survey and help them gather more data :)

    Seems to only cover smaller/independent artists.
    User-interaction oriented
    8Tracks http://8tracks.com/ Listeners are able to search through existing mixes
    and/or as create their personal mixes. 

    Users can search for mixes by individual artist,
    specific genre, or by utilizing the “cloud” feature
    that sorts mixes by clickable, emotionally and/or
    psychographically descriptive tags (i.e. autumn,
    Love, sad, eclectic).

    Users can “like” entire mixes or “star” individual
    tracks within them in order to facilitate quick access
    in the future, and can also “follow” other users,
    Effectively subscribing to the mixes they create.
    Everyone’s Mixtape http://everyonesmixtape.com/ Manually going through other mixes. http://www.youtube.com/watch?v=PAN13R3arnc
    This is My Jam http://www.thisismyjam.com/ Discover new music by “following” other users. 

    Users change their track of choice (jam) on a regular
    basis and their “followers” get to hear these new
    tracks that they introduce.
    From Music Blogs
    The Hype Machine http://hypem.com/popular See demo video http://www.youtube.com/watch?v=e9RNwdTuJX4
    Shuffler.fm http://shuffler.fm/ http://shuffler.fm/about
    General Recommenders
    Gnoosic http://www.gnoosic.com/ An AI system that engages in some kind of active
    Learning. 

    A part of the gnod.net website.

    Recommends music, books and movies & people.

    Not much of any other information available.
    TasteKid http://www.tastekid.com/ A general recommender system for music, books,
    Movies, TV shows, authors and games. 

    Not sure about underying technology.
    Maps
    Music Map http://www.music-map.com/
    Labyrinth of Genre http://static.echonest.com/LabyrinthOfGenre/GenreMaze.html
    MoMS
    (Map of Musical Styles)    		 http://static.echonest.com/playlist/moms/
    Other Links
    http://inspiredm.com/20-fantastic-ways-to-find-new-music-that-you-like-no-lastfm-pandora-inside/