What are Recommendation Systems?

Recommendation systems are an attempt to electronically provide recommendations to a user based on past experiences. In most situations, people get recommendations for things like TV shows, books and movies from their peers. If don't rely on their peers, they may check out critics web sites to see what the critic thought about the TV Show, book, or movie in question. The problem with taking recommendations from friends or from random critics, is that they are recommending you consumable content based on their tastes. While somebody may be your friend, their interests probably vary to some degree from yours and can dramatically effect whether you like the same thing or not.

One popular recommendation system that tries to deal with this issue is a technology called collaborative filtering. Collaborative filtering is a system that requires a certain level of input from you, then compares that data to the other users of the system. By comparing your past tastes, the theory is that we would be able to find somebody who has very similar tastes to you. Once we've found a series of people with very similar likes, we can recommend you movies based on what they think is awesome (and vice versa).

While this isn't 100% fool proof, as context isn't necessarily taken into consideration, the system is much less flawed than taking a random recommendation from a critic who may enjoy entirely different types of movies than you do.

Over the last five or six years, it has become almost trendy for College students to write their thesis on collaborative filtering and other recommendation systems. As a result, many different systems have popped up over the web. Big companies such as NetFlix have even put out a million dollar reward for the group that could come up with the system that came up with the best recommendations.

Of all the projects these are my personal favourite:

ApeFlix

ApeFlix takes your standard collaborative filtering system and adds a bunch of features the other applications don't have. If you have friends on ApeFlix, not only can you get personal recommendations for movies to check out, but you can also build a group. The system, will compare all of you, and find a list of movie that everybody in the group may want to see.

Jinni

Jinni is another large collaborative filtering web site. Jinni takes things a little bit further than other collaborative filtering systems and adds in context and plot style. By having these additional variables they are able to give you even more accurate recommendations

NetFlix

Netflix is probably the most widely used movie recommendation system because it is attached to the most popular movie and television streaming service available to date. The biggest flaw with their service is that it is based on a per account basis. So if you have multiple people in one home who all watch Netflix, it is going to be based on your entire household instead of on you personally.

Flixter

Flixter is part of Rotten Tomatoes, which is one of the biggest movie critic sites online. Flixter has a ton of users, so it has a great data set to work with. While they have some sort of recommendation system built in, from what I can gather, it is based on correlations between movies, and not based on tastes between users.

While I've focused on sites that provide recommendations based on movies, these same systems can be used on other things as well such as books and video games.