Search Term Returns

[Brizbee]

Hi folks, just thought I'd share these two search terms that returned totally inappropriate hits on my pictures recently, and wonder if anyone else has had a similar experience.

 

Search Term 1 - "bill wood"

 

Who's Bill Wood? No idea. But I know I don't have any pictures of him because I don't have any pictures of people full stop! This is the picture it returned -

 

 

An image of Wood Crane's-bill.

 

Search Term 2 - "dog lit up"

 

Yeah, not quite sure what that means myself. I assume it's a picture of dog lit by flash or a spotlight. Anyway it returned this picture -

 

 

An image of an Ox-eye Daisy, alternative name Dog Daisy, and this one happens to be back-lit.

 

Now, I'm trying my best to keep my captions and keywords relevant to each image and not add anything that doesn't apply. But this can't be great for Alamy's customers when they get results like this. And, of course, our CTR and hence our ranking take a hit as well. I think some serious tweaks to the search algorithm may be in order. What do you think?

[Allan Bell]

errrrm! well!  You are probably correct.

 

Allan

 

[Joseph Clemson]

Not a lot you can do about it, I fear. There is only one result for "Bill Wood", a sculpted piece in Knoxville. I assume this is what the searcher wanted.  Every other result is a because the search engine scrapes the bottom of the keyword and caption barrel to try and find some more relevant images. It is not often that so many irrelvant images come up in this way, but when there are few 'correct' results in the library, then some false positives are inevitable. It happens to us all and the bad occurances probably even themselves out over time across the contributor population.

 

The only contributors who will get more that their fair share of false positives are those who have filled their keywords with irrevelant or near irrelevant keywords. They will reap what they sow in that respect.

 

I would venture to suggest the back-lit Dog Daisy is probably falling foul of having just a little too much detail in the caption or keywords. I can't imagine any potential searcher looking specifically for a flower that is back-lit (though I'm not an expert in nature photography!). I would probably have left the 'back-lit' keyword out of the list, lest it does indeed produce false positives. Sometimes we can be a bit too detailed in keywording  - judging what is and what is not useful in that respect is something of an art and becomes easier the longer you have been doing it. Each morning I look at which of my images have appeared in searches unexpectedly and review them to see if I can rid myself of keywords which give rise to this kind of false positive. However, sometimes we just have to put up with them.

 

ETA. Nice images, by the way. You seem to keyword and caption very well and I think you are not a little unlucky to have been caught out with false positives  in this way. 

Edited October 18, 2019 by Joseph Clemson
Expansion of answer

[meanderingemu]

not sure what you want the search engine to do.  Your first image does have Bill and Wood in the data.  Why would you want the engine not to bring up an image that fits the request? 

 

i think a customer would then realise they have to refine their search, and then probably spend 2 more second and search again.  

[spacecadet]

15 minutes ago, meanderingemu said:

not sure what you want the search engine to do.  Your first image does have Bill and Wood in the data.  Why would you want the engine not to bring up an image that fits the request? 

 

i think a customer would then realise they have to refine their search, and then probably spend 2 more second and search again.  

What you can do about it is have phrase tags, where relevant, because they have higher search significance. "bill wood" as a phrase tag appears first in that search- it's the name of an artist, and repeated in the caption.

[meanderingemu]

14 minutes ago, spacecadet said:

What you can do about it is have phrase tags, where relevant, because they have higher search significance. "bill wood" as a phrase tag appears first in that search- it's the name of an artist, and repeated in the caption.

i agree.  If i was the customer, after Bill Wood gives me 12000 results, i would do "Bill Wood" which brings down to 91 results....  note that in both case, the search engine actually gave the only image featuring Bill Wood's work as the first result, so contrarily to the premise, i think the Engine did a pretty good job,

 

added, and then two images with a Bill Wood pictured ...

Edited October 18, 2019 by meanderingemu

[Brizbee]

28 minutes ago, Joseph Clemson said:

Not a lot you can do about it, I fear. There is only one result for "Bill Wood", a sculpted piece in Knoxville. I assume this is what the searcher wanted.  Every other result is a because the search engine scrapes the bottom of the keyword and caption barrel to try and find some more relevant images. It is not often that so many irrelvant images come up in this way, but when there are few 'correct' results in the library, then some false positives are inevitable. It happens to us all and the bad occurances probably even themselves out over time across the contributor population.

 

The only contributors who will get more that their fair share of false positives are those who have filled their keywords with irrevelant or near irrelevant keywords. They will reap what they sow in that respect.

 

I would venture to suggest the back-lit Dog Daisy is probably falling foul of having just a little too much detail in the caption or keywords. I can't imagine any potential searcher looking specifically for a flower that is back-lit (though I'm not an expert in nature photography!). I would probably have left the 'back-lit' keyword out of the list, lest it does indeed produce false positives. Sometimes we can be a bit too detailed in keywording  - judging what is and what is not useful in that respect is something of an art and becomes easier the longer you have been doing it. Each morning I look at which of my images have appeared in searches unexpectedly and review them to see if I can rid myself of keywords which give rise to this kind of false positive. However, sometimes we just have to put up with them.

 

ETA. Nice images, by the way. You seem to keyword and caption very well and I think you are not a little unlucky to have been caught out with false positives  in this way. 

 

28 minutes ago, meanderingemu said:

not sure what you want the search engine to do.  Your first image does have Bill and Wood in the data.  Why would you want the engine not to bring up an image that fits the request? 

 

i think a customer would then realise they have to refine their search, and then probably spend 2 more second and search again.  

 

Thanks for your replies.

 

I had a funny feeling there wasn't going to be a way of stopping these sort of results. What I would say is that the words "bill" and "lit" are parts of other terms, all be it hyphenated. I assume, therefore, that hyphens are ignored by the search algorithm.

[Joseph Clemson]

Correct, hyphens are ignored.

[meanderingemu]

1 minute ago, Brizbee said:

 

 

Thanks for your replies.

 

I had a funny feeling there wasn't going to be a way of stopping these sort of results. What I would say is that the words "bill" and "lit" are parts of other terms, all be it hyphenated. I assume, therefore, that hyphens are ignored by the search algorithm.

 

 

yeah, the hyphen just splits the word, but i am not sure any engine keeps them combined (having an hyphenated first name, i'm pretty sure I haven't come across systems that understand it's only one word

[meanderingemu]

1 minute ago, Joseph Clemson said:

Correct, hyphens are ignored.

 

technically not "ignored", because it splits components into all parts, so it gets replaced by a space in the search,

 

whereas apostrophe gets ignored and search for O'Reilly gives you all the  O'Reilly and Oreilly but not " O Reilly"