Thursday, March 27, 2008

Analysis for Mar 20 - Mar 27

This week, i carried on with my listening sessions, documenting the recommended artists, songs and my mood during the session. for the first time in all the sessions i have had to date, last.fm finally understood what i was looking for. or maybe it was just a coincidence. but this session was definitely better than all my other sessions. also the artist recommendations are getting really awesome now.

i would love to have a resident cognitive scientist help me understand the relation between my listening mood and ratings, but i guess that's asking for too much !!! you are welcome to download my excel spreadsheet file and do some analysis yourself. just dont forget to share your results with me !

A summary of my findings to date:

1. Songs played are directly related to the artists who are in the recommended list.
2. If an artist is removed from the recommended list, his songs no longer play as recommendations.
3. The song play list does NOT deviate from the artist recommendation list.

4. A track is scrobbled only after you listen to it for a few minutes/seconds.

5. Mood plays a big role in music choices.

6. Audioscrobbler generates recommendations on a weekly basis, which I think is influenced by the listening habits of my 'neighbors' and of course, myself. The popularity of an artist or track helps it to climb or fall in the recommendations list.
7. Audioscrobbler recommends audio tracks, albums, artists, events (concerts), videos and even users (neighbors).

8. I have a list of neighbors, who are people who listen to music similar to what I have been listening to ... last.fm calls them recommended users.

9. The list of artists which you may see on an artist page as being "similar" is based exactly on our user's listening habits. If a lot of users listen to Artist X, but also Artist Y and Z - Y and Z artists will become similar to X.

10. The recommendations are classified based on where they have been derived from (group recommendations, friend recommendations, and last.fm recommendations).

11. All the recommendations are tied up by Audioscrobbler to my listening preferences, but the recommendations vary very widely depending on the sources.

12. Recommendations from friends reflect their taste and I found most of these recommendations to be far from what I was looking for. But this is a social networking based technology, and listening to other people sometimes does open up interesting choices.

13. Neighborhood of friends!!! Yes we can actually see all the people who are listening to music that u have in your profile. This neighborhood helps to showcase 2 things:

a. Popularity of the music you are listening to.

b. Potential friends who you can add to your list.

14. There is also a recommendation list for albums and concerts and other music related events.

15. If the track length is less than 30 seconds, last.fm radio player refuses to scrobble the track.
Links:
Excel file documenting my listening sessions: http://www.filesend.net/download.php?f=842019c33f4638ee67c62c4eb54b10d1

Wednesday, March 19, 2008

Analysis for Mar 01 - Mar 19

After random and prolonged sessions of feeding my preferences into last.fm, it is finally time to document the recommendations last.fm throws at me. i am maintaining a excel sheet of all the recommendations, including the weekly recommendations of artists.

my first documented session was on March 17. i started off the session by documenting all the artists. then i let last.fm do its thing by letting it 'play my recommendations'.

Observations:
1. songs played are directly related to the artists who are in the recommended list.
2. if a artist is removed from the recommeded list, his songs no longer play as recommendations.
3. the song playlist does NOT deviate from the artist recommendation list

my first session threw up some pretty bad artist and songs. after removing some of these artist from the list, the song list deviated towards some really decent tracks. some of these tracks were infact ones i have heard before and are my all tiem favorites, and some were new tracks which were suprizingly awesome too!!!

but we all know music makers are not a predictable bunch of ppl. no matter how good an artist is, he or she or they always do release some music which might not fit on your scale of 'good' music. I love audioslave, and i love all thier songs, each and everyone of them. but some audioslave tracks are a lil too slow or grungy for me. you can attribute this to the fact that sometimes, listening to a artist perform a particular type of music programs your mind expect a certain tone, pitch and idealogy from that artist. any deviation by the artist from his original music style will put a lot of people off. because at the end of the day, we appreciate music, not the artist's ability in coming up with a totally radical piece of music, especially if you dont like that style of music. i would stop listening to audioslave if they were to get into country music ... for sure i would .. most Definately i would.

i know the statements above sound crude and judgemental, and i do not mean to generalize my views to all the music loving people of planet earth, but i have seen this happen one times too many. metallica, slipknot, ratm, chris connel. some ppl like eminem seem to pull off this transition with ease, going from rapping about issues plaguing our society and its understucture, to singing about money mercedes and fame.

moods. i know i have already documented thiis in one of my earlier posts, but when i started documenting my recommendations, i realized the critical role mood plays in shaping my recommendations. so i decided to document my mood too !!!

going back to my excel sheet, in my first session i ended up banning a lot of stations from my playlist. i also removed a lot of artists from my list. the second session proved to be a little better, with some decent tracks showing up in my list. by the third session i had a pretty good set of artists and songs being churned out from my last.fm radio.

i plan to now wait for a week before starting off another spate of listening sessions. hopefully i can determine some pattern in my tastes.

i will also talk more about pandora and how i believe music recommendation engines work towards satisfying listeners in my next post.

Link to my lat.fm recommendation excel sheet:
http://www.filesend.net/download.php?f=76ec86040ae994630f5b549db4481c23

Tuesday, March 11, 2008

Observation: Scrobbling

Back from a looooong break, i started up my last.fm player today looking for some new features to play with. i started off by creating a station with the tag audioslave (artist tag). the track like a stone came on - my favorite.

At the top of the last.fm player are 2 bars which look like a long progress bar. the top one states the current track the player is playing/scrobbling. the bar below talk about the station and the artist/song on which it is based.

A few seconds into the track, i noticed some activity in this top bar. at the same moment the track started playing, a blue progress bar started filling up the bar. a few seconds later the whole bar was blue, and then the words 'scrobbled' were flashed in the corner of the bar.

To sum it up, a track is scrobbled only after you listen to it for a few minutes/seconds. i will time this progress bar and get more info on it later. for now, i know that i have to actually listen to the track for a few seconds before it is scrobbled into my profile.


Update (March 12, 2008):
Link: http://www.last.fm/forum/34905/_/168133

The above link points to the fact that a track has to play for a predetermined amount of time, before it gets scrobbled and registered with the last.fm recommendation engine. though the source of this info cannot be validated, it is reassuring to know that someone else other than me has actually noticed this !!!!