The Googleplex Blog: Interestingness Is to Flickr As PageRank Is to Google Harold Davis’ over at the Googleplex Blog writes about Flickr’s interestingness and says that in a lot of ways the algorithm is very similar to Google’s Page Rank technology.
“”There are lots of things that make a photo ‘interesting’ (or not) in the Flickr. Where the clickthroughs are coming from; who comments on it and when; who marks it as a favorite; its tags and many more things which are constantly changing. Interestingness changes over time, as more and more fantastic photos and stories are added to Flickr.”
There’s an implication of recursiveness in this description. My photograph will have more interestingness if the people who view it, comment on it, and mark it as a personal favorite themselves have a high interestingness quota–which, in turn, can only be the case if these people have been favorited (etc) by those with relatively high interestingness.
The recursive nature of Flickr’s interestingness makes it analogous to Google’s PageRank–the complex site ranking system used by Google that ranks sites fundamentally using the PageRanks of the sites that provide links (“inbound links”) to the ranked site.”
Although this may make a certain amount of hypothetical sense, I think that the greater intensity focused on human filtering at Flickr works better than Google’s Page Rank algorithm. The problem is that Davis seems to be comparing web search to image search. Apples and Oranges. A better comparison is to simply compare Google’s Image Search results with Flickr’s Image Search results.
What you will find when you search Flickr’s image search results ranked by interestingness (which by the way guys, should be the default view) and compare them to Google Image Search results is that Flickr’s results are vastly superior. Google Page Rank may work well when dealing with text and blogs, but it does not work near as well when dealing with images.
A limitation of Flickr’s Image Search of course is that their universe is much smaller than the Google Images — but a couple of points on this.
1. Flickr is growing. Now with 1.5 million members their library will only get larger and larger. Still they should do more to incent the best photographers to post on their site. They should create a rewards program to recruit photographers as well as figure out ways that top photographers can monetize their work through Flickr — integrating Flickr search results into Yahoo! would be a start at minimum.
2. For most search results only the first two page search results matter anyways. What difference does it make if Google has 1,600,000 images for the term “love” and Flickr only has 1,600 (these are not actual numbers) if Flickr’s first two pages return superior images?
3. The strength of Google Image Search / Weakness of Flickr Image Search is in the comprehensiveness. Flickr needs to improve here. Flickr has all the best shots of cats and dogs and sunsets and flowers and San Francisco and New York that you’d ever want to see. They are not so good at finding obscure images that you still can find at Google Image Search even if the quality is worse. Sometimes a bad photo is better than no photo at all. Do a search for Irtish (an obscure river in Russia) and you get zero results from Flickr and 244 from Google Image Search.
Of course a way for Flickr to compete here would be to simply give interesting photos as ranked in flickr the highest page ranks on Yahoo! Image Search. Then Yahoo! image search would have the best of both worlds: a high degree of relevance for popular Flickr subject matter with the breadth that comes with the more automated algorithm that they are currently using for Yahoo! Image Search when Flickr has no images.
Seems pretty simple to me — but I repeat myself.