Signal Patterns Music Survey

Yesterday I found a link to a music recommendation function on Signal Patterns which is kind of neat - similar to MoodLogic. Daniel Levitin, PhD, a former mentor and friend I've since lost touch with, but has been and continues to be uber successful, participated in MoodLogic's founding and participates in the scientific board of Signal Patterns. His home page is here. I suspect that the music recommendation tool is at least partially his brainchild. It's a really good idea, and I suspect that for some people it works. I would encourage people to give it a try - it uses 40 song samples of about 15 seconds each and asks the user to rate on a scale of 1-9 whether the sample is liked or disliked. Let me know whether you like it or not.

The recommendation tool didn't work for me. I could take it again, that option is always available, however I suspect that part of the reason it did not work is that some song samples do not go anywhere: the meat of the song was not chosen as part of the song sample. If the meat of the song is absent, how can it be an accurate representation of the "music type" being evaluated? Signal Patterns uses Imeem and its social networking platform after the survey. Once I was switched to the Imeem site through Signal Patterns, I was asked to include my favorite artist. I don't really have a single favorite, but I wrote McCoy Tyner, and was then prompted to check off tracks of his that were favorites. None of the Tyner tracks listed were my favorites, so I backed out and entered Jethro Tull. This then led me to check off "Teacher", "Bouree" and "Locamotive Breath" as my favorite tracks (at least of those listed). (For the record, I also input Snake Oil Medicine Show as a favorite band and they were not found.) From these tracks and my evaluation from Signal Patterns I got the Imeem playlist that you can visit here. You can take a listen to the tunes that were recommended and see if those songs would be recommending the songs on this list. You can also take a gander at my Signal Patterns Music Survey Results below.
I'm really not sure whether the generated recommendations are determined from Signal Patterns (what I suspect) or are Imeem influenced (less likely), but the playlist doesn't work for me for the following reasons:

1) I've listened to the list and there is no new music - everything is already familiar, if by no other means, it is by popular names.
2) There are several songs by few artists; this cannot be all that exists!
3) I'm not keen on more than two or three of the songs.
4) I usually cannot stand ballads, sad songs or romantic songs, other than Sonatas. So much for these survey results...
5) I like up-tempo tunes, as a general rule (perhaps not the ones that were played, though).
6) I'm the king of complex tunes - I dig complex rhthyms, time signatures, layers, instrumentation.
7) Relaxing music is hard to identify - what really is relaxing? Easy listening?
On a positive note, I will offer that I'm usually not interested in lyrics, but would rather hear a nice hook or groove and someone justing singing "La la la"- the music survey may have gotten that right.

In general, all automated music recommendation tools I've evaluated are hit and miss (and probably more former than latter). Thusfar I still think Pandora is the best "automated" tool, although as I've said before it still has a lot to be desired, but it is at present successful. This tool by Signal Patterns is a neat idea, just like MoodLogic was, but I'm not convinced the approximately five minutes I used to take the survey were beneficial other than to give me something to write about today that didn't have anything to do with "Happy New Year". I may be a snob in this regard and this tool may be very valuable to others, afterall, Compete had Signal Patterns at ~16K unique users last month. That's pretty impressive. As alwasy I'd appreciate feedback on whether readers of this blog have better experiences with these music recommendation tools.

It is my sincere hope that Hounds to the Music, LLC will be able to add that "something" that is to be desired by music recommendation and discovery tools. We're working on it and anticipate that the launch will be this summer.

By the way, Imeem needs to significantly increase its bandwidth...I'm just saying.