pancakes

MicrostockGroup Sponsors


Author Topic: Best Match 2.0  (Read 33262 times)

0 Members and 1 Guest are viewing this topic.

bittersweet



lisafx

« Reply #1 on: December 09, 2008, 17:22 »
0
Thanks for the headsup.  Sounds ideal!  If it works it may be the best search results in the industry.

Only question I have is will non-exclusives still be at the back of the bus or do we get equal placement according to the applicable search terms?

shank_ali

« Reply #2 on: December 09, 2008, 17:30 »
0

bittersweet

« Reply #3 on: December 09, 2008, 17:31 »
0
Well, if they mean what they say, their primary goal is to get the best match in front of the buyer. This would mean the best keyword match based on the keyword ranking they've done. As it is now, the keywords may be all the same for images showing up, but for some they would not be as accurate as others, and other less relevant factors are weighing more heavily on the order of the results.

With the new system, it sounds like the boost will be given to those files which have actually been downloaded as a result of being found with that keyword.

I guess it will depend on how many files end up with the same statistics. It will be interesting to see how it shakes out. From the multiple references to "very different", I'd say that it might not shake out the way some (many) might expect it to.


bittersweet

« Reply #4 on: December 09, 2008, 17:32 »
0

« Reply #5 on: December 09, 2008, 17:43 »
0
Are we going to see this keyword ranking somewhere? So we can remove keywords that drags us down or replace them with better ones.

It looks like they are introducing something like Google's search engine optimization (SEO).

« Reply #6 on: December 09, 2008, 17:43 »
0
well either good or bad, it will be an interesting shaking of the box

vonkara

« Reply #7 on: December 09, 2008, 17:44 »
0
OK so we will need every brains to know how exactly it will work now... To start next week

To know like who will be on first page when every files will have 5 extremely relevent keywords
« Last Edit: December 09, 2008, 17:46 by Vonkara »

jsnover

« Reply #8 on: December 09, 2008, 17:49 »
0
Proof of the pudding as they say.

Let's see what happens once this is implemented, both to search results - do you get a nice mix of images, old and new, etc. - and to our sales.

Where will vectors end up? They've been trashed in the best match for a while now, so I hope they resurface (I'ld like to start back on that tack again, but have been contributing images only because it seemed a waste to dump vectors into a search that trashed them).

« Reply #9 on: December 09, 2008, 17:53 »
0
Quote
With the launch of our new keyword ranking system the culmination of years of work by thousands of people will come to fruition, providing iStock clients with what we think will be the most impressive and most exact search results, for community created media, found anywhere on the planet.

This kind of assertation scares me.  :)

Regards,
Adelaide

bittersweet

« Reply #10 on: December 09, 2008, 17:55 »
0
Proof of the pudding as they say.

Let's see what happens once this is implemented, both to search results - do you get a nice mix of images, old and new, etc. - and to our sales.

Where will vectors end up? They've been trashed in the best match for a while now, so I hope they resurface (I'ld like to start back on that tack again, but have been contributing images only because it seemed a waste to dump vectors into a search that trashed them).

It has been suggested (though not confirmed by anyone at all) that this latest best match was to allow those files for which data had not been collected to "rise" in order to rank the keywords. This would explain the older files being sold, and the most popular being pushed back. It does not, however, explain an across the board handicap assigned to non-exclusive files (as some have believed it happening).

Interesting theory anyway.

bittersweet

« Reply #11 on: December 09, 2008, 17:55 »
0
Quote
With the launch of our new keyword ranking system the culmination of years of work by thousands of people will come to fruition, providing iStock clients with what we think will be the most impressive and most exact search results, for community created media, found anywhere on the planet.

This kind of assertation scares me.  :)

Regards,
Adelaide

I think it's just a euphemism for "Please, please, God, let this work how it's supposed to!"
 :D


« Reply #12 on: December 09, 2008, 18:00 »
0
From a buyer view, will relieve some frustration. When I did active buying, the worst was getting images returned on search that were totally irrelevant. It just made me abandon searched much sooner.

From a submiiter viewpoint, weighting of each image based on it's own merit would be great. Let's hope it works.

« Reply #13 on: December 09, 2008, 18:01 »
0
I don't know how it will affect us yet but I am glad they are at least making a change(I hope it will work for most of us)fingers crossed!

traveler1116

« Reply #14 on: December 09, 2008, 18:03 »
0
The last two recent best match changes cut my sales by almost 50% both times, no optimism here.  But what can we do except wait and see.

« Reply #15 on: December 09, 2008, 18:07 »
0
Oh, lets not be so pessimistic. Maybe it will work OK. :)

« Reply #16 on: December 09, 2008, 18:16 »
0
This should be interesting.  Everytime Istock "wiki's" one of my files, sales on that file plummet there and increase elsewhere.  (And buyers are finding it using the deleted terms).  I don't see how the new best match will solve that.

bittersweet

« Reply #17 on: December 09, 2008, 18:29 »
0
A little more about the "blend":

Quote
Posted By rogermexico:
Quote
Posted By kelvinjay:
Quote
Posted By kkthompson:
On 2008-12-09 15:09:33, bitter wrote:

For artists it means a massive shift. The results for everyone, are going to be very, very different.


Given that massive changes to the best match are one of the main complaints in any best match thread, it will certainly be interesting to see how this one pans out. I guess we'll just have to wait & see...

Yeah, I know it's kind of a contradictory sounding message, but the goal is to have one last big shake so that we can move to something more stable. By blending this ranked keyword data into the algorithm, it should make it on the whole more stable. In the past we've been seeing really wild swings whenever the guys have gone into tweak the mix - this is aimed at alleviating those swings and giving us something more reliable.


« Reply #18 on: December 09, 2008, 18:32 »
0
A possible way to get non-biased people to accurately grade the quality of keywords would be to provide a "rate keywords" button with every image.

The rankings would have to be weighted according to how many images one has *downloaded* (purchased).  Those who only upload (sell) images would have zero weight attached to the rankings they submit.  Those who both download and upload would be weighted according to (number_of_downloads) minus (number_of_uploads), to prevent individuals or gangs of uploaders from buying a few images so that they can then vote themselves or their pals upwards.  To prevent contributors from establishing sock puppet downloading accounts, the length of time a downloader has been actively purchasing images would also have to be weighted in.  I.e., if you suddenly announced this feature, you would have to ignore rankings from contributors who suddenly create a new downloading account in order to give themselves a boost.

Hopefully customers who find that they have been victimized by keyword spam would be motivated to give the offending files a "zero (most keywords are spam)" rating, and when they find a good file they would give it a "five (all keywords are accurate)".  It would be easy to flag contributors who mysteriously receive a lot of zeroes/ones and a lot of four/fives, to check whether the high ratings are suspicious (e.g. originating from a small number of buyers).

Likewise, the image quality ratings should be strongly weighted towards *customer* opinions, and (net) contributor ratings should be either ignored or else given a completely different category.

Most images would probably not get any rank, so a contributor's "average" (per image) ranking would have to be used in the best match.  I.e. if they have uploaded 1000 images and 50 of them are ranked, then the average rank assigned to those 50 images is used to weight all of the contributors' photos in the best match.

If the rankings of image and keyword quality are more "real" (from the customers point of view, with the biggest customers having the biggest say) then they could be made a significant part of the best match.  These rankings would help turn the focus from "contributor oriented" and biased, towards "customer oriented".  In other words, let money talk and B.S. walk.

vonkara

« Reply #19 on: December 09, 2008, 18:33 »
0
I don't see how a search algorithm could see what is the subject in the image and look at the keywords if they are relevent... Then how else it could work if it's not by the one with less keywords. I don't know, I'm just asking??

I hope to see my best files in front this day, because my portfolio is actually in the best position to have sales since one of the latest best match change. Without saying that I have a good overall search result. But speeching of my own portfolio sorted by best match it was nice lately

A little more about the "blend":

Posted By kkthompson:
On 2008-12-09 15:09:33, bitter wrote:

For artists it means a massive shift. The results for everyone, are going to be very, very different.

OMG now I'm lost  ::)
« Last Edit: December 09, 2008, 18:35 by Vonkara »

bittersweet

« Reply #20 on: December 09, 2008, 18:49 »
0
I don't see how a search algorithm could see what is the subject in the image and look at the keywords if they are relevent...

It can't see the image, but it can tell if an image was purchased as a result of a search with that keyword. If it was, then it means it is likely a relevant keyword. (There will be some exceptions surely, but overall I'd think this would be more accurate than not.)

« Reply #21 on: December 09, 2008, 18:57 »
0
My interpretation of keyword ranking is a little different, so it will be interesting to see how the system is actually implemented. I see it as the order of a keyword in a list (and maybe the number of words in the list) determines how much weight it has, not some arbitrary? weighting determined by istock. But maybe I'm wrong

« Reply #22 on: December 09, 2008, 19:19 »
0
I get the feeling we will all have to go through every keyword again, like we did with disambiguation.

« Reply #23 on: December 09, 2008, 19:20 »
0
Does this mean???? that if an image for example: a picture of a woman sitting in a chair on the grass.
Say it has 3 keywords "woman, chair, grass"
A buyer searches for the keyword "woman" and buys the image. Now the image will rank higher under the keyword "woman" but will not rank higher under the keywords "grass and chair". Another happy shopper buys the image by searching keywords "woman, chair" now image moves up best match for "woman and chair" but not for keyword "grass". Individual keyword ranking would be great as long as the image is not penalized for having the relevant keyword "grass".

« Reply #24 on: December 09, 2008, 19:20 »
0
Maybe they will (finally) also evaluate relevance by searching title and description.

Regards,
Adelaide


 

Related Topics

  Subject / Started by Replies Last post
what is up with the Best Match?!?

Started by traveler1116 « 1 2  All » iStockPhoto.com

31 Replies
8850 Views
Last post March 17, 2011, 13:28
by Sedge
114 Replies
24881 Views
Last post April 03, 2011, 11:51
by BaldricksTrousers
235 Replies
44458 Views
Last post April 09, 2011, 17:30
by Sadstock
4 Replies
5095 Views
Last post May 27, 2011, 20:36
by Sadstock
1 Replies
2611 Views
Last post December 21, 2017, 10:35
by christiano

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors

3100 Posing Cards Bundle