Comments on: Help! My iPod thinks I’m emo – Part 1

By: Finding New Fans With Recommendation Systems | Grind EFX

Finding New Fans With Recommendation Systems | Grind EFX — Mon, 29 Jun 2009 12:02:37 +0000

[...] day, as the team that reached the goal in the Netflix Prize illustrate. But, as Paul Lamere put it, recommendation systems are broken. The most commonly-used recommendation engine (collaborative filtering) still keeps people stuck in [...]

By: Spotify + Echo Nest == w00t! « New Business Concepts Music Industry 2009

Spotify + Echo Nest == w00t! « New Business Concepts Music Industry 2009 — Thu, 21 May 2009 14:52:12 +0000

[...] ballads. There are a boatload of other issues with wisdom of the crowds recommendations – I’ve written about them previously, suffice it to say that it is a challenge to get a CF-based recommender to give you good [...]

By: Spotify + Echo Nest == w00t! « Music Machinery

Spotify + Echo Nest == w00t! « Music Machinery — Tue, 19 May 2009 17:39:59 +0000

[...] There are a boatload of other issues with wisdom of the crowds recommendations – I’ve written about them previously, suffice it to say that it is a challenge to get a CF-based recommender to give you good [...]

By: OlegG

OlegG — Sat, 18 Apr 2009 06:15:26 +0000

Hi Paul,
I do not listen to a music very often and I have a big trouble when I am trying to buy a new disk: a choice is so huge, but my budget is pretty limited.
I suppose that some experts should create a structure of music categories big enough to reflect most of people tastes.
After that (I am very sure) people would love to use them.
As a software developer I can provide a lot of ideas and algorithms to support such activity.
I am absolutely sure that a usage of just a 100 criteria will be able to produce billions and billions music categories.
Is it so difficult to find out 100 criteria to categorize a music?

By: bcurtu

bcurtu — Sun, 12 Apr 2009 16:02:00 +0000

I have implemented a events recommender for dooplan, using mainly CF algortihms, but also Content Based algorithms in order to avoid problems like novelty problem…

When the second part of this article for? it’s really well explained.

By: Adondai

Adondai — Thu, 02 Apr 2009 12:23:46 +0000

If you take a look at some other blogs you’ll see most have a seperate spot where the trackbacks go – either above or below the normal comments – and usually consisting of just one line.

Their messy and distrupt the conversation that go on in the comments.

Disqus is the easiest solution =P It has good spam filtering as well.

By: plamere

plamere — Thu, 02 Apr 2009 10:46:27 +0000

I agree that there is lots of data out there already that would make recommendation much easier. But it is difficult to share this data – it is hard to reconcile users (my last.fm name and my pandora name are different), and most of the APIs only allow for non-commercial use.

By: plamere

plamere — Thu, 02 Apr 2009 10:44:46 +0000

Not sure what you mean. Is there a problem?

By: Adondai

Adondai — Thu, 02 Apr 2009 02:44:28 +0000

Btw can you do something with the trackbacks on this blog?

By: Adondai

Adondai — Thu, 02 Apr 2009 02:40:13 +0000

Hey – couldn’t agree more with your thoughts.

However what I don’t understand is there are already dozens of metrics and APIs which if a music recommender used, should be able to do a lot of what you suggest?

I don’t know anything about programming, but all of these music recommenders seem to try and create the perfect engine themselves when any of them could use information provided by all these other sites? You posted about new services and APIs on your old site all the time…