Max's Output

Archive for March 2011

How to Build your Personal Trusted Web for Spam-free Search

with 2 comments

“Google is flooded with spam”, that is what we hear all around. It is sort of true. For some searches there are 9 out of the 10 first entries are spam. On the other hand there is a spam-free zone – it is the stream of links from your friends in social networks. If you follow people more or less selectively you will never see spam in your news feed. There is actually no way for spammers to get into your streams unless you follow them.

We are working on combining these two worlds: spam-free zones of social networks and Web search. The basic idea is simple. You sign in with your accounts (such as Twitter, Facebook, etc), we extract links posted by your friends and build the list of web sites that is your trusted Web. You can then limit your search to your trusted Web. We are experimenting with this idea in various ways. For example, the links from your friends can be extended with links from friends of friends (i.e. one hop via social graph). Or it can be extended with pages which are linked from the pages posted by your friends (i.e. one hop via Web graph). We can also compute popularity of Web sites among your friends and use it to improve the ranking of search results. We are going to leverage The Tweeted Times technology to build it.

Note that it is not only about spam-free search. It is also a great personalization tool as you search through sites that are the most relevant to you. You get these sites on top of search results that rarely happens when you search over the whole Web.

To reduce search results to the trusted Web we use Blekko API. Blekko is a really cool search engine that allows you to “slash” the Web with a list of Web sites. We generate Blekko slashtags that you can use to search only through your trusted Web. Combining our slashtags with other tags you can personalize it further.

There are ongoing efforts to improve search via social graph. Google already ranks higher and highlights those links in the results that your friends posted on social networks. Blekko allows you to search through Web pages liked by your friends on Facebook (use /likes slashtag). But in both cases it is about indexing links to particular page. In its current form Google and Blekko work more like a bookmarking system to find pages linked by your friends. We want to take it further – to go from page links to Web sites as described above. We will keep you posted on the progress.

Written by maxgrinev

March 15, 2011 at 7:23 pm

Posted in Uncategorized