Tuesday, July 21, 2009

Idea #22 Personalized Recommendations Through Twitter Stream


If you use Twitter a lot, the words steadily coming out from your Tweets, aka, your Twitter stream, contains a lot of information about you. It tells about your interests, location, food you eat, events you go to, opinions you have, and more. These are extremely valuable data to make highly personalized recommendations if well analysed. Most importantly, they are free and publicly accessible.

In fact companies have started mining these data streams in hope of providing highly differentiated services. For instance, moonit.com uses your Twitter behaviors, Tweets, followers and followees, to suggest who you should date. Your Tweets already tells who you are, why should you fill out a lengthy phycological survey ?

I would like an algorithm to recommend me the following based on my Tweets :
  • Events happening around me that are interesting to me (LikeMe.net ? No, still based on survey questionnaires)
  • People with like minds who I should meet and partner with
  • Restaurants and Cafes with just my tastes (yelp.com ? not quite.)
  • Cool gadgets that I should check out
  • News and websites that I will like (Stumbleupon.com? well maybe.)
  • and much more

These are just some simple examples. The opportunity is endless, as long as a data mining algorithm can be built to discover the associations between keywords, and cluster keywords into categories, as talked about earlier, the power of association.

BTW, you should try out the extreamly cool way to visualize your Twitter stream, Portwiture.com, which expresses your Tweets with Flickr photos.

photo credit Konstantin Sutyagin.

No comments:

Post a Comment