Jump to content

Undiscerning keyword proximity results


Recommended Posts

I am despairing of ever improving my CTR as long as Alamy's keyword proximity search engine continues to produce nonsensical results.

 

My latest list of views included two returned for the search Used Belt [WOP]. Only 240 images were viewed and two of them were irrelevant and mine!

 

One example image is CMCRBA  CMCRBA.jpg

 

The words Used and Belt appear nowhere in the keyword boxes. The caption (an Alamy live news image) includes the sentences '...the subject of a planning application to convert the 17th Century grade II listed building to eight dwelling places, plus 21 new houses built on surrounding green belt land. The land taken will include a large car park currently used by visitors, including school coach trips...' (my emboldening for clarity).

 

There is no sensible reason this should ever be returned in a search for used belt, even on the basis of a proximity search.  This kind of result both plays havoc with my CTR and gives the customer the impression that the Alamy collection is badly keyworded, neither of which is desirable.

 

This is by no means an isolated example, I seem to to encounter this issue almost every day. Am I the only one? Is the only option to reduce all my captions and keywords to a bare minimum that brook no possible misinterpretation by the search engine? Or is it time this was addressed by Alamy so their search produces really meaningful results? 

Link to comment
Share on other sites

I notice a few of these, but it is only a few. I haven't analysed it for a few years as it doesn't seem to affect sales, but IIRC the percentage of single result searches, and those are the problem, was in the low single figures. I correct it when I can but I don't worry about it.

Link to comment
Share on other sites

I am despairing of ever improving my CTR as long as Alamy's keyword proximity search engine continues to produce nonsensical results.

 

My latest list of views included two returned for the search Used Belt [WOP]. Only 240 images were viewed and two of them were irrelevant and mine!

 

One example image is CMCRBA  CMCRBA.jpg

 

The words Used and Belt appear nowhere in the keyword boxes. The caption (an Alamy live news image) includes the sentences '...the subject of a planning application to convert the 17th Century grade II listed building to eight dwelling places, plus 21 new houses built on surrounding green belt land. The land taken will include a large car park currently used by visitors, including school coach trips...' (my emboldening for clarity).

 

There is no sensible reason this should ever be returned in a search for used belt, even on the basis of a proximity search.  This kind of result both plays havoc with my CTR and gives the customer the impression that the Alamy collection is badly keyworded, neither of which is desirable.

 

This is by no means an isolated example, I seem to to encounter this issue almost every day. Am I the only one? Is the only option to reduce all my captions and keywords to a bare minimum that brook no possible misinterpretation by the search engine? Or is it time this was addressed by Alamy so their search produces really meaningful results? 

 

The solution looks quite simple: after the news images have gone into the main collection, remove the caption to the description field and write new captions.

The problem with that is that the description field only shows on a large screen, when the image is horizontal and when there are only two lines of keywords. And even then only the header and one or two lines of description are visible.

I wish the description field would follow right under the caption.

 

wim

Link to comment
Share on other sites

Thanks for raising this which has made me realise an error in my thinking as I had thought that searches were influenced by the keywords and not so the caption. Presumably, from Wim's suggestion, searches don't use the description field?  

Link to comment
Share on other sites

This particular example was a news image but the problem is by no means limited to news images with long captions. On the same day I had an irrelevant, non-news,  image viewed in a search for used cars uk. The word 'used' appeared in the caption only. The word 'cars' was the last of the words in the main keywords section. It is a nonsense that an image keyworded in this way is returned in the first 240 images in a search for used cars.  It should no be necessary to have to prune captions and keywords to such a minimum as to eliminate false results like this!

 

To add insult to injury, the search failed to return three images of mine which had used cars in the essential keywords field, while managing to return five irrelevant images where the words were widely separated in the caption and main keyword fields. The more I look at the matter, the more I am at a loss to understand what is going on. 

Link to comment
Share on other sites

Thanks for raising this which has made me realise an error in my thinking as I had thought that searches were influenced by the keywords and not so the caption. Presumably, from Wim's suggestion, searches don't use the description field?  

 

Yes. Description field has not been included in searches for years now. Initially it was.

 

wim

Link to comment
Share on other sites

 

This kind of result both plays havoc with my CTR

 

 

...and everyone else's.

 

Alan

 

Broadly speaking you are correct - everyone's CTR will be impacted in the same way, though a little more so for those like myself who routinely caption and keyword in a comprehensive manner. I could lean towards brevity, at the expense of clarity in some cases, but it is my argument that if keyword proximity search is going to be Alamy's major search strategy, it really needs to be a lot smarter than it currently is.

Link to comment
Share on other sites

 

Thanks for raising this which has made me realise an error in my thinking as I had thought that searches were influenced by the keywords and not so the caption. Presumably, from Wim's suggestion, searches don't use the description field?  

 

Yes. Description field has not been included in searches for years now. Initially it was.

 

wim

 

I just had a search result which included words which were ONLY in the description field (I hardly ever use that field), nowhere else.

Link to comment
Share on other sites

it is my argument that if keyword proximity search is going to be Alamy's major search strategy, it really needs to be a lot smarter than it currently is.

 

 

Yes, but proximity searching should only affect the order of the matches, not the total. If the words 'used belt' are searched then it's perfectly correct that all images containing those two words should appear in the results, but where the words are not in proximity they should appear lower down the order, and this is exactly what appears to be happening in this case.

 

I would argue that this is less detrimental to your CTR than, say, a search for Seaton turning up pictures of Seaton Delaval, which is at the opposite end of the country, and which if Alamy implemented quotes properly would not show up (but see below...).

 

I've realised recently that even if Alamy does implemented quoted keyphrases, to make them effective they would need to remove captions from search matching, otherwise the caption would also have to show two-word names in quotes, which would look rather odd. Actually I think this is a good idea anyway - to me it makes sense that searches should be done on keywords and nothing else.

 

Alan

Link to comment
Share on other sites

I just had a search result which included words which were ONLY in the description field (I hardly ever use that field), nowhere else.

 

 

 

Are you sure? That shouldn't be possible.

 

Alan

Link to comment
Share on other sites

 

I just had a search result which included words which were ONLY in the description field (I hardly ever use that field), nowhere else.

 

 

 

Are you sure? That shouldn't be possible.

 

Alan

 

+1

 

Kumar

Link to comment
Share on other sites

 

it is my argument that if keyword proximity search is going to be Alamy's major search strategy, it really needs to be a lot smarter than it currently is.

 

 

Yes, but proximity searching should only affect the order of the matches, not the total. If the words 'used belt' are searched then it's perfectly correct that all images containing those two words should appear in the results, but where the words are not in proximity they should appear lower down the order, and this is exactly what appears to be happening in this case...

 

....

 

Alan

 

I can't detemine anything about the way images were ordered in in the used belt example, other than to re-affirm my point that images where keywords are widely separated and only found in the caption field, really ought not to be found at all in such a search. The minimum requirement ought to be that at least one of the search terms is in the essential or main keyword field.

 

However. I can say for sure that in my second example, used cars, irrelevant images of mine with those two words widely separated in the caption and main keyword fields appeared earlier in the search, ahead of other images of mine where the same keywords are actually in the essential keywords box. I know these far more relevant  images were further down the search because the customer gave up at 240 images and my relevant images never even got viewed. 

 

As far as I can see, the keyword proximity search system is not working as it ought.

Link to comment
Share on other sites

 

it is my argument that if keyword proximity search is going to be Alamy's major search strategy, it really needs to be a lot smarter than it currently is.

 

 

...I would argue that this is less detrimental to your CTR than, say, a search for Seaton turning up pictures of Seaton Delaval, which is at the opposite end of the country, and which if Alamy implemented quotes properly would not show up (but see below...).

 

I've realised recently that even if Alamy does implemented quoted keyphrases, to make them effective they would need to remove captions from search matching, otherwise the caption would also have to show two-word names in quotes, which would look rather odd. Actually I think this is a good idea anyway - to me it makes sense that searches should be done on keywords and nothing else.

 

Alan

 

 

I wish they would implement quotes properly as it would prevent  ambiguous search results which seem to me to be inevitable if one relies primarily on proximity keywording.

Link to comment
Share on other sites

 

I just had a search result which included words which were ONLY in the description field (I hardly ever use that field), nowhere else.

 

 

 

Are you sure? That shouldn't be possible.

 

Alan

 

I deleted the description field, it was the only place the words were. I'll try to check, but I'm sure.

 

EDIT: My apologies, I must have been mistaken as I just tried some searches with other words used only in my descriptions and didn't find anything, so it may have been one which I had already purged a few days before.

Link to comment
Share on other sites

 

it is my argument that if keyword proximity search is going to be Alamy's major search strategy, it really needs to be a lot smarter than it currently is.

 

 

Yes, but proximity searching should only affect the order of the matches, not the total. If the words 'used belt' are searched then it's perfectly correct that all images containing those two words should appear in the results, but where the words are not in proximity they should appear lower down the order, and this is exactly what appears to be happening in this case.

 

I would argue that this is less detrimental to your CTR than, say, a search for Seaton turning up pictures of Seaton Delaval, which is at the opposite end of the country, and which if Alamy implemented quotes properly would not show up (but see below...).

 

I've realised recently that even if Alamy does implemented quoted keyphrases, to make them effective they would need to remove captions from search matching, otherwise the caption would also have to show two-word names in quotes, which would look rather odd. Actually I think this is a good idea anyway - to me it makes sense that searches should be done on keywords and nothing else.

 

Alan

 

 

As much as I agree with your "used bell" case (and other people's examples) on the face of it, it's still something of an oversimplification.  You have to consider how to deal with things like "used car" and "second hand car" (and maybe "secondhand car", "used automobile", "second hand automobile", "secondhand automobile"... I can go on and on and on and on).  To do it efficiently you really have to process the key words at the point of indexing, rather than search.  I can think of dozens of ways in which Alamy could hugely improve the relevance of their search results.  I do wonder if contributors trying to "game" the system might be the limiting factor... along with cost, of course.  My main point here is that there are some obvious fixes which solve some problems but ultimately this needs ground-up analysis to prevent a further raft of unintended consequences.

 

[i should say that I only quoted you to make the use of your example]

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.