Creating a simple recommendation system for reddit: any advice?

I want to create a simple user-based recommendation system for reddit, mainly for personal use. The idea is to get training data in the "liked" and "disliked" items for the user (this requires the password, so training will be on one user only). The recommender will then periodically retrieve new links from reddit and present the most promising ones to the user.

So far the features I've identified are the following: url domain, number of upvotes, number of downvotes, author id, and link title.

I don't really know what family of machine learning algorithms I should use. What would you recommend?

Thanks for your help!

EDIT: if you want to try your hand at it, here is a simple python script that retrieves the data from reddit.