Topic on Talk:ORES

Patterns based on sessions from anonymous users?

2
QuimGil (talkcontribs)

(This is a question emerging from a discussion in Catalan Wikipedia.)

Imagine this situation: an anonymous (IP) user makes 5 edits during one session. They are all subtle vandalism, introducing wrong words and concepts that require certain knowledge to detect (i.e. that Town X is not in the coast).

Now imagine, that from these five edits, three are detected and reverted, but the other two still remain in place. It could be that the other two edits are legitimate, but chances are that the other two edits done by the same user in the same session are vandalism too, just not detected.

Is ORES analyzing this type of situations already? If not, is this a pattern that could be considered?

EpochFail (talkcontribs)

ORES currently doesn't analyze beyond a single edit. This is something we're hoping to look into thought. For Phab:T155756, we're planning to take a whole session of edits and build features from them. Once we have that infra, we can experiment with other models too.

Reply to "Patterns based on sessions from anonymous users?"