We are exploring some options for automatic detection of files uploaded from somewhere else. This bot will only run on Commons, on files where the source is "own work", and where the uploader is not "trusted" (to be determined what that means, exactly)
Approach to detect possible copyright violations:
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T134802 Improve the curator workflow for reviewing new files | |||
| Open | None | T120453 Copyright violation detection tool for Commons | |||
| Duplicate | None | T123517 Automatically check Commons uploads for possible copyright violations | |||
| Duplicate | None | T230561 Create a ML model to score new files in commons for copyvio issues |
We are exploring some options for automatic detection of files uploaded from somewhere else.
@MarkTraceur: Has that exploration happened in the last five months, and what was the outcome?
If not, does anyone plan to work on this soon (or should the priority be changed?
Thanks!
@Aklapper Sorry about that, yeah, this is on the back burner because we have no good leads on partnerships with services that could provide image checking for us. Google has said in no uncertain terms that they don't have an open API for this, and I think partnerships are being negotiated with some other system, but are not finalized. In any case, it's blocked internally and low priority for us currently.
I suggest different approach that is independent on external service (this may be compementary way):
@Halfak @Ladsgroup is it something thatalready planned in ORES?
Not necessarily a check suitable for all uploads, but comparing a new upload against a hash for an 'office' actioned removal might be useful. You could compute an SHA-1 on a new upload and compare it against ones previously removed. (Like is done in ehcking for duplicate uploads). This could prevent accidental re-upload of previously removed material, with a suitable warning to the uploader.
Closing as the Multimedia team was disbanded in 2019, this task is largely similar, if not identical to the CWS one, and this task was created one year after T120453.