__ Continuation of http://www.microstockgroup.com/microstock-news/zymmetrical-user-stole-my-images/msg87149/?topicseen#new

It happens everywhere and yes, it seems that the majority are coming from Lynx for some odd reason. Maybe it's a robot translator.

Without beating you about the head over the usual inaccuracies in keywords. The first three with the girl, all have automobile in the keywords, and there is an auto somewhere in the image.  Grin The problem is that if someone does a search for 20 images, they get better matches, when it goes up to 40 or 80, Zymmetrical puts the LAST pages first, so lower matches are showing at the top of the search page, best matches show at the bottom. Strange?

I wasn't just complaining for the sake of causing trouble. If you have less images now, and you tighten the limits to start with, it's not going to be as hard to get the site under control as if you wait for a year and 1 million more images later to edit.

No stethoscope isn't a doctor, just like a pipe wrench isn't a plumber. A road sign showing a moose crossing isn't an automobile, and neither is a stretch or highway with no cars on it. How many buyers will get frustrated from too many vague matches and implied connections and go to iStock or Veer where CV is reigning in a runaway proliferation of mismatched key words. I suppose a poor selection of search words will produce poor results, (as in my example of one word "automobile) it's going to happen, but a front end loader, kid on a bicycle, tire and fuel gauge are not automobiles.

One way to get free help is add an option to, flag an image for bad keywords? Now you don't have to have someone going through all the images looking for obvious errors like a lady elephant where someone may have just bulk edited their keywords by accident.

I am in agreement, a stethoscope should not come up near the beginning for a  search on Doctor - it is however thematically related, and considering our relatively small inventory and strict editorial guidelines, there may be a balance achieved by throwing in thematically similar images. A good salesman can present new ideas to the customer if it's appropriate and make a deal that would not have happened had he not presented some variations of the customers request, right?

The keyword flagging tool may be an option but I can sleep ok with at least our internal keywording reviews in their current form - our Reviewers see everything and pay the same attention to the photo technical details as with the keywords, and will not hesitate to prune or reject outright for bad keywords.   If you look at www.zymmetrical.com/art/photos/ which are the latest additions, newest-first, the recent quality is decent.   Again the strange results you have seen that are/were outright wrong is isolated to the Lynx batch and just a temporary thing as the system blew a fuse because that 'user' had 125K photos when doing a routine update.

p.s. I don't deny some of the other issues you mentioned, we will see if they stick around when the index settles down. There is always room for improvement.

No demands, just questions and suggestions. 

It would be worth testing keyword flagging at least. If we found our staff getting into extended philosophical debates about why people decide particular keywords should be there or not, we run into a real scalability issue. If it would end up being mostly used in good faith for serious errors, why not. 

No, the fileid's are tied to the root category pages. We already have to worry about dupe content penalties enough with the search engines having mostly the same content on 9 website mirrors (zymmetrical.co.uk etc.), moving the files around the hierarchy of the site would be asking for trouble.  I'd be interested in seeing what you produce, if it can be packaged as widget or such it could possibly be added to our http://www.zymmetrical.com/info/labs/.