Hello, Gordon, I'm not quite sure if I get your point well. Perhaps it's because of my poor english, trying to concentrate on language I lose overall meaning. So maybe in exchange I'll just try to explain my view on case.
The entire idea is based on concept of conncections and their strength. Connections can be between everything - artists, songs, tags, groups etc. in any combination. Based on these connections we can get for example artists similar to current artist, or artists common for particular group. This could be done simply by taking some of strongest connections. My idea on how to create these connections and define their strength is by giving points. For example, when somebody listens to a song, connection to a song which he played before gets 4 points, connection to two songs ago gets +2, and 3 and 4 songs ago get +1. Numbers could be different, but the whole idea reflects the fact that when someone listens to some song, the next one will also be in his taste. Maybe points shouldn't be based on how many plays passed (for example 1 or 2 songs earlier), but a time that passed from playing the previous song (for example points = 20 / x, where x - how many minutes ago the song was played). There are many possibilities, maybe you'll come up with some more :-) Moreover, the points could be given in many other cases, and to different items. For example - when somebody plays a song of some artist, connection with the artist he played before gets some points. Or when a user adds an artist to favourites, connections to his other favourite artists get points. Or when user is attending an event, connections between his favourite artists and artists performing on that event get some points. And another idea is that each time points get to high (in typical database you can only store values lower than some limit), we could divide points of this type by 2. It has this advantage, that we get "dynamics" then. Older connections lose their meaning. What do you all think about that? I'd be grateful for any comments. thanks Maciek _______________________________________________ Libre-fm mailing list [email protected] http://lists.autonomo.us/mailman/listinfo/libre-fm
