User:Multichill/Categorization bot

Idea for a new categorization bot.

The current bot (imagerecat.py) depends on Commonsense. This is often slow or even worse: Broken. The new bot should be stand alone.

At the moment about 40.000 files are uncategorized, but used at some project: query and result. Would be nice if a bot could categorize these files.

The should work on a list of images (for example the the list of used uncategorized files). For each image:

• See where the image is used.
• For each article, see if it contains a commonscat link
• If the article doesn't contain a Commonscat link, go work on the categories and interwiki's of the article
• The process should be recursive with finding a link to Commons as the stop condition.
• There should be some sort of scoring
  • The earlier, the higher the score
  • Multiple hits for the same category is a combined higher score
• There should be some kind of filtering
  • Blacklist
  • Skip hidden categories
  • (etc etc)