All content in spam?
These reflections pushed me to a constant increase in the number of topics in my RSS-reader. On Habré, too, are constantly emerging articles in the style of "but I do not like topics about XXX on the main, give the opportunity to filter." As a solution, they propose to filter unloved words via Regexps in Yahoo Pipes.
The Bayesian algorithm is successfully used to filter spam. It is simple, learning and effective (cuts off up to 95-97% of spam). So why not use it to filter the flow of information?
Suppose that all the topics in the reader is spam. User behavior is no different from the usual - he reads the topic for the topic, noting to himself what he liked and what he did not like. Let's add one additional opportunity to the reader - to mark the topics you like (for learning the Bayesian filter). Then, after a period of training, the filter will be able to filter out topics that the user will most likely like. And put them, for example, in the "Read first" section.
')
You can go further and filter the information using a browser plugin.
Can move the topic to the blog "I'm crazy"?
The Bayesian algorithm is successfully used to filter spam. It is simple, learning and effective (cuts off up to 95-97% of spam). So why not use it to filter the flow of information?
Suppose that all the topics in the reader is spam. User behavior is no different from the usual - he reads the topic for the topic, noting to himself what he liked and what he did not like. Let's add one additional opportunity to the reader - to mark the topics you like (for learning the Bayesian filter). Then, after a period of training, the filter will be able to filter out topics that the user will most likely like. And put them, for example, in the "Read first" section.
')
You can go further and filter the information using a browser plugin.
Can move the topic to the blog "I'm crazy"?
Source: https://habr.com/ru/post/37213/
All Articles