Over the weekend I had some problem with causality. I found this page on Commons,
http://commons.wikimedia.org/wiki/File:Pine_Grove_Springs_Hotel,_Spofford_La...
which claims to have gotten an image from my web site,
http://ookaboo.com/o/pictures/picture/1849742/Pine_Grove_Springs_Hotel_Spoff...
which in turn claims to have gotten the image from
http://en.wikipedia.org/wiki/File:Pine_Grove_Springs_Hotel,_Spofford_Lake,_C...
which also claims to have gotten the picture from Ookaboo. Obviously this doesn't make sense. Although this image is old enough that we're not going to have anyone getting angry about attribution, but our users deserve to have sensible provenance information for this image.
Ookaboo almost certainly got this image from Wikimedia Commons by an automated process. It seems that somehow, Wikimedia Commons re-imported the image from Ookaboo.
I'll grant that a large majority of images on Ookaboo were gathered by an automated process that found images on Commons and enriched the metadata. However, Ookaboo now allows users to add images, and Ookaboo is now gathering images that aren't in Commons, see:
http://about.ookaboo.com/a/about/add_a_picture
Specifically, we're tuning our image loading process to fill gaps in what Commons has, so I'd like to invite people to add our images to Commons -- however, we ought to have a mechanism to prevent mistakes like the one above.
Any ideas?
On 28 July 2011 17:52, Paul Houle paul@ontology2.com wrote:
Specifically, we're tuning our image loading process to fill gaps in what Commons has, so I'd like to invite people to add our images to Commons -- however, we ought to have a mechanism to prevent mistakes like the one above. Any ideas?
There are open-source "similar picture" matchers. Running them over the entire image archive might be an, ah, interesting experience.
- d.
On Thu, Jul 28, 2011 at 12:25 PM, David Gerard dgerard@gmail.com wrote:
There are open-source "similar picture" matchers. Running them over the entire image archive might be an, ah, interesting experience.
Open source or not, TinEye and Google's "search by image" work just fine.
On 7/28/2011 1:25 PM, David Gerard wrote:
There are open-source "similar picture" matchers. Running them over the entire image archive might be an, ah, interesting experience.
Ookaboo itself has the problem that it sometimes ingests the same image or almost the same image twice. I've been thinking about running a GIST filter or something like that on the images, partially to find the dupes but also to create a new navigational axis. It's not a high priority because I need to get the user engagement up and also take more advantage of the semantic data that I've already got.
On Thu, Jul 28, 2011 at 6:52 PM, Paul Houle paul@ontology2.com wrote:
Over the weekend I had some problem with causality. I found thispage on Commons,
http://commons.wikimedia.org/wiki/File:Pine_Grove_Springs_Hotel,_Spofford_La...
which claims to have gotten an image from my web site,http://ookaboo.com/o/pictures/picture/1849742/Pine_Grove_Springs_Hotel_Spoff...
which in turn claims to have gotten the image fromhttp://en.wikipedia.org/wiki/File:Pine_Grove_Springs_Hotel,_Spofford_Lake,_C...
which also claims to have gotten the picture from Ookaboo.Obviously this doesn't make sense. Although this image is old enough that we're not going to have anyone getting angry about attribution, but our users deserve to have sensible provenance information for this image.
Ookaboo almost certainly got this image from Wikimedia Commons byan automated process. It seems that somehow, Wikimedia Commons re-imported the image from Ookaboo.
No, that's not the case. The picture was uploaded on commons from en:, where in turn it was uploaded by the specified user, who claims without reason to doubt it that it was scanned by him from an out-of-copyright source. Only several weeks _after_ it was copied from en: to commons the reference to ookaboo.com was added.