Machine Learning
In the latest episode of The Mac Developer Roundtable, Andy Matuschak talks about machine learning as a new direction for user interfaces. One of the examples he gives of a good use of it would to tell him in the morning which of the 200 unread articles he should read in the few minutes he has before leaving for class. This is a really good use of machine learning and one I’m happy to say Cyndicate can handle for him (or anyone). The functionality can be achieved by combining two of Cyndicate’s features, automatic ratings and smart folders.
Automatic ratings takes care of the machine learning part of the equation. As you rate articles while reading, Cyndicate learns what you like and don’t like. Then when new articles are received Cyndicate will give it a rating based on your history. Think of it as your email spam filter in reverse. Instead of trying to decide what it spam, it tried to figure out what you would be interested in. The system works really well once you get it trained.
Smart folders are the other part of the equation. Cyndicate’s smart folders include rules to filter based on the article’s rating. So if you make a smart folder with rules to show unread articles and articles with a five star (and maybe four star) ratings you will achieve Andy’s example. Every morning you’ll have a list of four or five star articles to quickly read through that is based on what you’ve liked in the past.