3 Comments

Summary:

When you get book recommendations from Amazon or music suggestions from Last.fm, they’re based not just on your own shopping and surfing history, but on the preferences of people who like the same things you do. It’s called collaborative filtering, and it’s not always an ideal […]

matchmine logoWhen you get book recommendations from Amazon or music suggestions from Last.fm, they’re based not just on your own shopping and surfing history, but on the preferences of people who like the same things you do. It’s called collaborative filtering, and it’s not always an ideal way to find new things to read, watch, or listen to.

Suggestions based on collaborative filtering can be helpful, but they can also leave you wondering where your ability to think for yourself went. One antidote can be found in the personalization approach taken by matchmine – it profiles your likes and dislikes on numerous levels, then matches that multidimensional taste profile with the characteristics of specific movies, blogs and other content.

[poll]

How it works

KeyForge rating music

The system revolves around the MatchKey, a feature in which users describe what they enjoy. The MatchKey is then stored on your hard drive, where it can be accessed by content personalization applications that help refine matchmine’s understanding of you.

You create your MatchKey using an online application called KeyForge, which starts by recording your “demographic archetype” made up of zip code, year of birth, and gender. Then it walks you through the four types of currently supported content: movies, video, music and blogs; you either rate specific examples (in the case of movies, video and music) or tell it what categories you like (in the case of blogs).

You then save your MatchKey onto your computer, where it can be read by locally-installed Adobe AIR applications and used to make recommendations. Right now, matchmine offers two beta applications: MyMovieMatch, to help you find movies, and MyGumballMachine, which makes recommendations from all four categories.

Why it might be better

What I like about matchmine is that it lines up my multidimensional taste profile (my MatchKey) with the multidimensional profile of a piece of content. For example, I like “The Princess Bride” because it combines comedy with romance and a bit of fantasy; matchmine can find other movies that have similarly specific profiles.

matchmine reminds me a bit of Pandora, which doesn’t use collaborative filtering but rather searches for music based on the characteristics of music you say you like. Pandora, however, doesn’t construct a personal profile of you to match to the music; it starts with music you specify. [Do you like Pandora or Last.fm? Take our poll at Web Worker Daily.]

matchmine can also match you to other people, by computing the similarity of your respective MatchKeys. That would be another path to find content you might like.

matchmine works across content types and services, effectively bypassing the compartmentalization of personalized recommendations. But it does so in a way that doesn’t compromise privacy, because you retain control of your MatchKey. Plus, when used to make recommendations, it’s not associated with any identifying data.

Future plans

In January, matchmine will roll out integration with partners Fuzz (for music) and Filmcrave (for movies). Peer-to-peer online trading platform Peerflix has been named as a partner as well. CEO Mike Troiano says the next wave of partners will include both small discovery players and larger content providers.

Early next year, matchmine will offer the capability to store MatchKeys on a server, making the download optional. Then people can come across it as a feature on some web-based content service, discover how it works, and then use it on another one, all without having to understand what it offers in absence of a specific use. It’s hard to understand in the abstract, as Troiano admits.

If matchmine can get MatchKeys incorporated into various content platforms and show users the benefits of truly unique personalization without the need to compromise their privacy, this could be a winner.

  1. [...] December 6, 2007 by miketrap To be honest, I never imagined I’d be so excited for us to appear in somebody’s blog. If you’re going to be in one, though, this is the one. [...]

    Share
  2. [...] Matchmine was featured on GigaOM recently. Very interesting recommendation engine that avoids the majority of privacy issues that [...]

    Share
  3. [...] That’s all I can say for now (veil of secrecy descends). Read more about Matchmine at GigaOM. addthis_url = [...]

    Share

Comments have been disabled for this post