Monkey News
Among the few dozen (about 60 or so) news feeds I get are the Â鶹¹ÙÍøÊ×Ò³Èë¿Ú feeds for world, science, and related news. The problem is that I was reading so many feeds that I spent more time sifting through news I have already read in one form only to see it in another. Finding the hot topic at any one time was difficult and almost defeated the purpose of the system.
What I wound up doing was designing a system that will determine the hot topics on the fly and lump related stories together, ranking the whole thing by popularity. It's updated every two hours, and tracks breaking news quite efficiently. By cutting across so many sources from around the world, I get a more complete picture that I would find with only one or two sources, and I can scale to numerous sites with ease. unlike bayesian techniques, there is no training to do, as the approach automatically determines categories.
At this point, monkey news has been running for almost two years without little problem. It needs some more polish, search features, and improve navigability, along with better story correlation through a real scoring function.
Comments