When google announced a preview of google image swirl I was compelled to check it out. After all lifeblob is also in the business of photo discovery and it is always good to know what other companies are doing in this area.
The preview is neatly built in flash and presents a very interesting visualization. This reminds me of a similiar visualization they did earlier for a “wonder wheel” view of navigating through search results. The part that I did not like though was that exploration ends very abruptly – perhaps this is because they are running on a limited data set and will not be a problem in the final version.
That brings me to the question – what is google trying to achieve through this? Is it just a ploy to counter the advances of microsoft in visual search or is there a bigger agenda behind this?
My verdict: There is a bigger agenda here. Read On.
1) Searching through images is not an easy task and given that google’s key focus is search, they would want to establish a big lead in the image search market as well. So far, google image search attempts to identify duplicate images so that they do not present duplicate results to the user. With this kind of a visualization however, google lets you explore variants of the same image so that you have more freedom to choose the one that fits your bill.
2) This is an excellent way to provide more visual detail around a searched topic. Imagine a wonder wheel kind of icon next to your regular search results, clicking on which opens up the image swirl widget and lets you see photos related to a search result.
3) Search is intent driven – you type in a few keywords and then choose a result that matches your intent. Google has used this information very effectively for its regular search to verify that the results being shown match the user intent and also to improve its data mining algos to continuously improve the results.
Images suffer from a terrible handicap that they don’t have any metadata to organize them and google typically relies on information around the images, tools like image labeler and some level of image analysis to understand what they are about. By presenting the results in a visualization like this and watching user interaction patterns, google will get excellent information about images and to improve their image search.
4) Navigating through images this way is very enjoyable and gets people to spend more time on google properties. Now, that may not necessarily align with their standard way of measuring search effectiveness ( its counterintuitive, but google wants you to spend as little time as possible on its search pages, which is their metric for measuring how good their results are ) but hey, the more time they spend here, the more time they are off Microsoft’s visual search and thats a good thing :). Plus, the metrics that apply for text based search may not apply for visual media search anyway.
What do you think about this?