Has the possibility of adding Baynesian filtering to articles been discussed? I’m new to newsblur (after the google reader announcement).
Many sites I follow have a wide range of topics, only some of which I care about. I’m currently using the keyword matching filtering to some good affect, but would love to get baynesian filtering of articles - that’s a good deal more powerful for this kind of thing.
I built bayesian classifiers four years ago and it was a disaster. Not only is it naive (which the current system also is), but it doesn’t show you why it selected a story or hid another story. It was also pretty slow. I decided against it and went the explicit filtering route. Much happier, since all it took was a better trainer, which took ages to build right.
Fair enough. Thanks for the response!
@samuel, Did you ever look at cream or at fever. My experience with cream is that bayesian filtering really does work quite well.