Forums » General Discussion Search

Are Personal Catalog ratings relative or absolute? New Reply

Author Post
Posts: 50
Registered: Mar 10, 2010

Do you guys have a universal metric for what each rating implies (e.g. 8 = strongly likes), or is it a purely relative scale (e.g. artist1 = 3 and artist2 = 8, therefore user likes artist2 more)?

I want to add ratings to a personal catalog, in order to influence dynamic radio. So of course, this only matters if ratings are taken into account. I read earlier that this was coming. Has it been implemented yet?

Cheers,

Peter

Posts: 1075
Registered: Sep 08, 2008

Hi Peter - when we use ratings in taste profiles and playlisting we use a rating scale centered at 5. So anything above a 5 is considered a positive rating and anything less than 5 is a negative rating. A '0' is the worst rating, '10' is the best. And yes, ratings are taken into account with catalog and catalog-radio playlists. -- Paul

Posts: 50
Registered: Mar 10, 2010

Thanks.

So how does this compare with artists who aren't in the catalog? Is an artist with a rating of 5 more likely to appear in a catalog-seeded radio vs an artist not in the catalog? How about an artist with a rating of 4? Trying to figure out the nuances so I can get the right balance.

Also, I assume that an artist without a rating is the same as a rating of 5?

Posts: 1075
Registered: Sep 08, 2008

doppledoer - there are lots of moving parts here. When scheduling a playlist for each candidate song we calculate a 'user affinity' which is a value corresponding to how much we think the user will like the song. A bunch of data goes into this value including artist and song ratings, plays, skips, bans, favorites etc. We also calculate a 'user familiarity' for each candidate song. This is a value that corresponds to how familiarity the particular listener is with the item. This is based upon plays, skips, ratings, and so on. When we schedule songs for the playlist we take into account the user affinity and the user familiarity combined with the distribution, variety and adventurousness settings for the playlist to pick the best songs. So to answer your specific question, if the only piece of data we have from a user about a particular artist is a rating of '5', that artist will have a neutral 'user affinity' score, but it does get an increased 'user familiarity' score, since by rating that artist, we at least know that the user knows about the artist. Depending on the adventurousness setting, we may steer toward this artist (if adventurousness is low) or away from this artist (if adventurousness is high).

Hope this helps

Paul

Posts: 50
Registered: Mar 10, 2010

Definitely helps. Thanks

Reply to this Thread

You must log in to post a reply.