How Does Learning Work?

NewsBrain learns when you correct any of its Interest Level predictions. It does this by propagating the correction through your Interest Profile, adjusting the Interest Levels of relevant topics so that future predictions will be more accurate.

NewsBrain predicts your interest in articles that it fetches, and in topics in your Interest Profile. It predicts article interest by looking at profile topics relevant to the article, and taking into account the interest levels of those topics. When you correct NewsBrain's interest prediction on the Articles Screen or Article Screen, NewsBrain adjusts the interest levels of the relevant topics. The next time NewsBrain makes a prediction on a similar article, it uses the new topic interest levels to make a more accurate prediction.

NewsBrain also predicts your interest in topics listed in your Interest Profile. In addition to adjusting topics when you set an article's interest level, it adjusts related topics when you set the interest level on a profile topic itself - which you can do from an Interest View on the Topics Screen, Topic Info Screen, or Relevant Topics Screen. Your adjustment, to a lesser extent, ripples through topics both above and below it in the profile tree. So if you adjust the interest level of the "American League" topic, lower level topics like "New York Yankees" are adjusted, as are higher level ones like "baseball" and "sports".