Machine-Learning RSS Reader based on simple user feedback

I've created an RSS Reader that uses machine learning to deconstruct articles marked by the user as liked/disliked, scoring article's elements (words, categories, authors...) and allowing the user to filter out articles irrelevant to their interests. Users can also order articles by score instead of date, showing the most relevant ones at the top of the list. I'd be interested to know whether anyone here would like to try this system once it's online? Perhaps it could help researchers, skimming lots of articles on their subject every day, or even ordinary people, trying to find a car of their dreams, filtering out the noise of online auctions irrelevant to them. I'm aware that there's some RSS readers that already partially do this, such as NewsBlur, Feeder or Inoreader but they either over-complicate this by requiring users to set up lots of filters manually or to play an RSS-reader-training-cookie-clicker by having them to rate each valid/invalid word, one by one. I'd prefer avoiding comments about RSS decline, if possible please.