Collectif L'OUVRE-BOITE

Exemple à propos de la musique : on pourrait prendre d'autres exemples, dans la littérature, ou encore autre part. Va pour la musique ... Lisez cet article : Online music sources : finding tracks that matter. Un extrait est présenté ci-dessous. Sans doute connaissiez vous déjà le "très ancien" LivePlasma, qui fait un riche usage de la cartographie. Connaissiez vous les autres services présentés plus bas ? Beaucoup de systèmes de recommandation utilisent, comme il se doit, le filtrage collaboratif. Certains utilisent aussi des tags. Ce qui est intéressant est l'intégration (Web Services) de contenus textuels divers, en provenance par exemple de iTunes Store, d'Amazon, de Wikipedia. Mieux encore, l'intégration de contenus musicaux. Et lorsqu'on peut partager (au sujet d'un artiste, d'une composition, d'un tag ...) c'est encore mieux.

J'ai brièvement essayé GenieLab. Assez bonne intégration, pas encore de cartographie, partage possible, tags embryonnaires. Le service semble développé par le seul Robert Rose. Voudriez vous essayer les autres services, et rendre brièvement compte ? De la synthèse qui en résulterait, on pourrait tirer les principaux traits (attraits) d'une maquette ouvre-boîte à réaliser. Il s'agirait d'aller nettement plus loin (notamment dans l'intégration) que la série actuelle des "36 prototypes". Le but de la maquette serait, comme d'habitude, d'explorer les usages, et d'aider à définir un modèle économique. Peut être pourrions nous associer à cette maquette (disponibilité de contenus à intégrer, partenariat commercial ou bénévole, etc.) des producteurs et distributeurs de contenus musicaux et textuels ?

With Big Radio on the way out, where do you turn for music recommendations?

Is your music collection in need of an update? Have you scoured every corner of the iTunes Music Store only to be left wanting more? With the decline of broadcast radio and everyone jacked into their iPods, how is anyone supposed to discover good new bands these days? As with many modern-day dilemmas, several good solutions can be found on the Internet.

The web offers a variety of resources for those searching for new music to listen to. Websites connect fans with bands and their music. Encyclopedic online directories catalog musicians and their works. Other websites allow you to compare your favorite artists to others’ favorites in order to receive personalize music recommendations.

Online Retailers

Many online music retailers offer their own recommendations. In addition to the wealth of data contained in various playlists, the iTunes Music Store includes cross-referenced sections titled “Influencers” and “Contemporaries.” Listings also include a section titled “Listeners also bought” to help guide you to similar offerings.

On Amazon.com, product pages include a section titled “Explore Similar Artists.” Amazon also lists more personalized recommendations based on your buying habits, in a section called “Recommended for You.”

Recommendation Systems

Another group of sites has sprung up solely to recommend new music using a technique called collaborative filtering to make personalized recommendations based on music tastes. By gathering your likes and dislikes and comparing them to the data collected from other users, sites such as Musicmobs, Audioscrobbler, and UpTo11.net are able to provide some of the most unique and insightful recommendations available.

An interesting feature common to several of the recommendation systems are dynamically generated maps of artist similarities. LivePlasma has some of the most information-rich of these maps. Colorful spheres represent each artist, with the size of the sphere indicating their relative popularity. Artists are grouped and connected by lines to indicate similarities.

Other sites with similarity maps include Gnoosic, GenieLab, and an off-shoot of Audioscrobbler, called Audioscrobbler Browser. Clicking on different artists will reorient the map and load a new group of similar artists. Browsing these maps can give you a more thorough understanding of the similarities between specific artists.

Each recommendation site offers it’s own unique set of features. GenieLab band listings include background information culled from various other websites, such as Amazon.com and Wikipedia, as well as links to MSN Music and the iTunes Music Store for buying the music you like. MusicStrands includes links to purchase songs from Amazon, Disco Web, the iTunes Music Store, and Walmart.

For more granular recommendations, both Soundflavor and MusicStrands each focus their suggestions on songs and playlists, rather than artists. Soundflavor also adds additional analysis, by digitally examining the actual sound of songs in its database.

Each of the recommendation sites also has it’s own unique algorithm for determining similarities. A search for the same band will likely result in differing recommendations from each site. For instance, LivePlasma tends to group and suggest artists in closely related genres, while Gnossis seems to suggest the most obscure and least obvious recommendations.

Human Touch

As with any technological solution, don’t be too quick to dismiss the human equivalent. While your friends may not share your exact musical taste, they can still be an excellent resource to tap into, especially if you want to try something far outside your normal comfort zone.

I’ve found many of my favorite artists by getting recommendations from my music-loving friends. Don’t hesitate to ask your friends for recommendations; and perhaps most importantly, don’t forget to spread the word when you find an artist that you think others will like.

-------------------------------------

Mots clés et tags :

cartographie cartography recommandation recommendation filtrage collaboratif collaborative filtering Web Services partage sharing tags musique music intégration integration agrégation aggregation contenu content usages uses modèle économique business model