Google joins the Social Search Party

With the release of Co-op, Google has begun to allow its users to improve Google’s search results. Wink fundamentally believes that incorporating user input can improve search relevancy. The Wink Social Search Engine is designed to give ordinary people the power to improve search.

Alan Eustace at Google agrees. In her post Google machines vs. humans, Bambi Francisco wrote:

“When I asked Google CEO Eric Schmidt what was the most exciting product launch in the last year, he said ‘Google Co-op.’ In his words: ‘It’s a powerful idea’ because it gets people to help Google structure the data. Through a co-op, ‘user-generated data becomes part of the answer,’ Schmidt said. I think this is a brilliant idea for Google to tap into the more arcane searches. Among the digerati, these searches are called the ‘long tail.’ Getting to the long tail of searches, the results of the most convoluted or obscure search queries, is something that humans can help with. ‘Machine algorithms aren’t good at it,’ Alan Eustace, senior vice president of engineering at Google, said to me. … ‘For the time being, the human judgment is still much better’”.

We’re excited to hear that Google thinks that “humans” have a valuable and intelligent contribution to make to search. (although I’d rather call them “people”).

On the surface this seems like a huge admission for Google, the company that pioneered today’s machine powered search that leads the market. But it’s not really that odd to say that people are better at knowing what’s relevant. It’s hard for machines to predict what people’s intentions are. For more about the whole idea of a database of intentions, check out John Battelle’s blog. When somebody types a query term they are submitting the words they think best describe the information they seek.

Say someone enters: “weight loss”. He isn’t looking for “pages that have a high density of occurrence of the string ‘weight loss’” nor is he looking for pages that have a high density of link flux (inbound links with the anchor text “weight loss”). He is looking for pages *about* the topic ‘weight loss’. Not surprisingly, this is the phrase used by others to describe such pages. Google results are usually relevant because they contain lots of pages that *contain* the term, and are *about* the topic. However, the Weight Watchers home page doesn’t use the words “weight loss” any where on it, so it doesn’t show up in Google’s results, whereas it is a top site that users have tagged “weight loss”, and therefore shows up in a search for “weight loss” at Wink.

So How Does Google Co-op Work?
As written by Saul Hansell in the New York Times article “Google Shows New Services in Battle of Search Engines”

“Marissa Mayer, Google’s vice president for search products, said in an interview that a chief purpose of the new product was to help improve Google’s main search by determining which sites gathered the most interest from users.”

Philipp Lenssen provided an awesome write-up on How to use Google Co-op.

There are three social aspects to Google Co-op:

Search Topics
Contributors create specialized searches for a Topic by specifying web pages/sites that they consider authoritative for that topic. They annotate each specified website by uploading “annotation files” (either tab delimited or XML) that associate “labels” (tags) with URL. The examples Google gives suggest that labels be used to specify things like audience, source type, document type, or subject. When a user subscribes to the topic, these annotations will show up above the search result.

Subscribed Links
Service providers can submit XML files that add their services into Google search results as subscribed links. These files are called “context files”. Users subscribe to the Subscriber Links for a provider they are interested in. Then when they do a search that matches an entry by the service provider, a “one box” is displayed above the search results containing information from that provider. A handful of services had subscribed links in the directory at launch. So far there doesn’t appear to be a clearly stated criteria for inclusion in the directory.

User Pages
There is talk of user pages where a user can collect together links on a Topic, but I haven’t been able to find where these are or how a user creates one. (Anyone? Anyone?) I assume they are supposed to be something like Wink Collections or Jeteye Jetpaks.
The Google Co-op still seems to be less about regular people selecting what shows up in search results through ordinary interactions, such as tagging (A good post on Tagging: what are the personal benefits of tagging part 1) bookmarking, etc. and more about publishers and webmasters submitting their content to Google searches via the one box for users who subscribe to them.

Wink on the other hand, is focused on delivering search results that are directly influenced by ordinary people during their searching experience rather results refined by publishers via XML uploads, etc. (”end users” rather than Web masters) To date, Wink has processed over 20 Million user generated tags and used them to deliver better search results. Wink’s PeopleRank and TagRank weave user input and tags respectively, into the search process. Try searching any term you would on Google. In many cases, Wink delivers a very accurate result. And if neither Wink, nor Google get it right - with Wink a mouse click is all it takes to set the record straight. If the top results aren’t relevant for your search demote them, if the truly useful link is on the second page - then promote it, find a link you like, tag it for future reference and if you see any search-spam - block it!

There’s a Wink Collection started on the topic of Google Coop. Check it for additional information about Google Co-op, and if you find anything new, you can add to it simply by logging in.

And please subscribe to the Wink Subscriber Link at Google Co-op. It’ll allow you to get people powered results from Wink for some of the most popular queries, while you are searching Google.
Cheers,
Michael

Tags: Googlecoop, Google Co-op, Wink, search, socialsearch, wink.com

22 May 2006 | observations | Comments