Bayesian Twitter Client
Summary
Gathering signal from the noise. Classify tweets automatically by learning what the user is interested in.
Criteria for Positive Input
- Clicking on a link in a tweet
- Replying to a tweet
- Favoriting a tweet
Inputs to Bayesian Filter
- Username of the tweeter
- Number of people mentioned in the tweet
- Number of links in the tweet
- Is the tweet a RT?
- Number of letters in the tweet other than links
- Expand the links so that words in the URL are counted
Other ranking factors
- Search for "congrats" or "so happy for you" to find significant changes in people's lives. suggested by @caseorganic
- Look for an unusual volume of replies in your timeline to specific people to see what's grabbing your friends' attentions.
Comments
