|Performance of your classifier|
|Number of examples in this session|
In just a few minutes you can have your own customized that classifier that will learn as you go.
- Use it to decide whether posts in your blog are relevant to your theme.
- Use it to decide what articles you would like to read from Hacker News (personalization).
Try to pick a concept related to pages. Something that is determined from the text. Here are some tips:
- The categorizer answers exactly 1 binary question. Is this page X? where X is a topic. X should be a cohesive topic--about one thing in the world.
- The topic should be a determined from the text on the page, not the layout, structure, video, or ajax requests.
- Good Topics: Finance pages, Science articles, Fluff news, Adult, Spam, Astrophysics, Pages you like.
- Bad: Pages with no text, pages with video / media, Concepts related to structure: such as find the slow-loading pages, find pages which crash, etc.
Enter URLs and provide feedback to the classifier as it tries to learn what you want. When you're done, you can use the api to continue scoring and training your model on your own site.