Tuesday, May 22nd, 2007

Differences Between Google’s ‘Universal Search’ and ‘Image Search’

There's been a fair amount of buzz lately around Google's 'Universal Search' - where they embed highlights from image, video and local search into the results for regular text-based (web) search. Since (a medical illustration site we administer) ranks pretty well on image search, I've been looking at how the image search results compare to the images shown in universal search. What I'm seeing is a bit odd...

Let's start with the image search results for "Down Syndrome"...

Snippet from Image Search for 'Down Syndrome'

The images shown above are just a snippet - they're images 1, 2, 3, and 6, 7, 8 in the results. Notice that the Netter image for down sydrome ranks #3.

Now take a look at the universal search results for "Down Syndrome"...

Universal Search Results for 'Down Syndrome'

You would think the images they pick to show in universal search results would be the top three from image search since "most relevant" shouldn't change based on whether the search is an image search or a web/universal search, but that's not the case at all...

The first image shown in the universal search results looks like a pretty normal baby - while the kid may have down syndrome, I wouldn't call it a good example of down syndrom. Plus, the image can't be found at all in the image search results (at least not in the first 200 images). Furthermore, when you click through to it you find the page no longer exists on the server. Google implies in some of their blog posts that if you opt-in on enhanced image search then your images are more likely to show up in universal search results, but I can't see how the image of the baby would ever get tagged with the phrase 'down syndrome'... Strike one for universal search...

The second image they present in universal search is of a guy on a treadmill. Even when you click through and look at the full-sized image, the phrase 'down syndrome' doesn't jump out at you - I'd guess he was an overweight Latino. The page the image is on is not about down syndrome, though the alt tag of the image does identify the guy on the treadmill as having down syndrome. Like the first image, it's nowhere to be found in the first 200 images in image search. Again, it's safe to say it's not one of the top three most relevant images on the web for down syndrome. (Strike two...)

The third image they present in the universal search results is of a doll with down syndrome. The thing is, you can understand the relationship to down sydrome based on the caption in image search, but there's no caption in universal search and based on the thumbnail you're left wondering why you're being shown a picture of Cabbage Patch Dolls when you typed in 'down sydrome'. This image does show up in the image search results (as image #6), and I would say it's relevant once you take a closer look at the image or see a caption, but without the caption I wouldn't say it's of the three most relevant images for down syndrome and its inclusion without a caption is just confusing. (Strike three...)

A picture of a perfectly healthy looking baby, a picture of what looks like an overweight Latino on a treadmill, and a picture of cabbage patch dolls just doesn't cut it for the search term 'down sydrome'...

Universal search is incredibly important to Google's future. They're smart people and they will get it right. But right now they'd do better if they just used the first three images from the image search results.

Tags: , , , ,
Categories: Google, Image Search

Leave a Reply