One of the richest areas of experimentation in search right now is how to rank real-time results. For the most part, that means finding relevance in Twitter and bringing up the most important Tweets for any given keyword (see OneRiot, Collecta,Scoopler). Today, real-time search engine CrowdEye is introducing its own real-time ranking algorithm called CrowdRank. It’s supposed to be like Google’s PageRank, but for the crowd.
Right now,real-time search is Twitter search because that is the richest source of real-time data. And Twitter search is essentially a form of people search. Twitter’s own search engine simply brings back a reverse-chronological list of the most recent Tweets that match the keyword you enter.
CrowdEye does that as well because often in real-time search you just want to see what is happening at this second. But now CrowdEye will let you sort by relevance as well, rearranging results by the most influential people on Twitter. (See screenshots below)
What exactly goes into CrowdRank? CrowdEye founder Ken Moss, who previously was a search guru at Microsoft, won’t reveal all the factors. But the number of followers someone has seems to be the main one. He says:
CrowdEye Rank has many inputs, and the list will be changing over time as we work to refine the algorithm. Obviously it includes things like how many followers you have and whether you are a “verified” twitter account. Less obviously are some factors we use to penalize spammers.
Fortunately, he includes other measures of influence too, like how many times any particular message has been retweeted. Otherwise @aplusk is going to show up at the top of every search.
But now that every person on Twitter has a CrowdRank, when CrowdEye returns results, it shows an actual CrowdRank number between 1 and 100 at the bottom right of each avatar for the top Tweets in results. There is also a directory of the top CrowdRanked Twitter users, but these seem to match up closely to the list of people with the most followers (which again brings us back to to @aplusk problem).
For any given search, CrowdEye returns the top Tweets as well as the top links. Another change today is that if you sign into CrowdEye with your Twitter account, you can follow anybody who comes up in search results or retweet a message without leaving CrowdEye. CrowdEye will also now give you a personalized list of people to follow based partly on who you are already following.
This list is much better. For me it suggested my former Fortune colleague David Kirkpatrick and New York Times reporter Brad Stone (I swear, I thought I was already following you guys—no wait, that’s on Facebook). It also suggests Stocktwits (I’m not really a trader), author Tim Ferris (yes), and MC Hammer (why not?).
And most ambitious of all, CrowdEye will create a personalized homepage showing you links and Tweets tailored for you (see bottom screenshot). It shows you the most Tweeted articles from your favorite pre-selected blogs and news sites or ones which match saved queries. So instead of an empty search box, you are greeted with a bunch of recent content to explore as filtered by both your personal preferences and the collective wisdom (or idiocy) of Twitter.
Stumpedia’s Social Rank algorithm uses your social graph to determine the relevancy of search results in real-time.
finally a twitter search / twitrank type mashup – surprised it took so long for this to appear, but it’s a great first step.
I did not know thats how they ranked twitter. Thank you for the info.
Their logo looks familiar!
Yearite!
Ya that logo is weak. Come on guys – be original. Lots of ideas left out there…
@Byrne – yeah…arrington designed it for them when he was in temporary retirement
Notorious Twitter spam is “World Of Warcraft”
Top Results by relevance – All Spam.
What is “most influential people” @solarpowerwell seems to make the list a lot, and what is he, an Affiliate Marketing Spammer with 7K followers. Not only is he first he is at the top 15 times!
Who is next on the list after all the spam? @BestNetContent relevant, after him, more spammers.
Influential People Seems to be “has most followers”, and I assume this is the only way to do it algorithmically.
This search needs to not just show most influential people on top, but “unique” most influential people. The most resent post from the most influential people.
I think they have something good going on here, but they need not give most popular the full result list, even if it was relevant content, it does not make for good results.
It reminds me of the old search engine traps from the late 90s when sites like Excite would end up with 5 pages of results from the same web site.
I hate it when spammers get through the holes, and this logic they have lets them in.
OK yea, playing with it more, this thing is learning live I think? This may be something done very well, wait and see.
It’s interesting, but IMHO search is bound to fail and is not adapted to real-time data. How can you search for something that, by definition, you’re not aware of (yet) ?
The key to the real-time web is discovery, not search. That’s what we attempt to do.
But do you really need Twitter to do that? Sure, it makes things easy… but why not define specific sources and then use NLP and keyword techniques to find out what’s most interesting?
Their design is basically a rip of Google Finance.
“CrowdEye will create a personalized homepage showing you links and Tweets tailored for you”
Something like that can be a social-bookmaking killer!
Real-time search is simply when content finds you, rather than you actively searching it out. Each user needs a way to control that stream.
Twitter is just a conglomeration of people rehashing content that is already available elsewhere – i.e. CNN, TMZ, TechCrunch, etc
This should be done through the use of natural language processing and keyword swarming. When MJ died, the number of content sources mentioning keywords such as MJ, Michael+Jackson, Michael, Jackson, etc was exceptionally high.
Sure you might end up with multiple keywords linking to the same content, but the point is that it rose to the top.
Looking at the company logo/colors … I initially thought it was a product from TechCrunch
Hmmm. Certainly an interesting space, but this service is not ready for prime time. In my testing, results were dominated by spammer accounts, and the service crashed more often than it yielded results.
Looking for popular twitter users who tweet about your favorite topics? Try feedmil.com – a real-time feed search engine
Any ranking system that uses incoming links/followers as a way to determine relevance is short sighted and discriminating
its look like google web site design…
communication, one to another is the important
means task one can do.
I don’t get it: what the interest of searching from a database where basically are saved the “answer to the simple question what are you doing now?”.
Let me rephrase: the interest would be obvious if the question was not answered by spammers…
Not sure that Crowdrank can test Pagerank!
Interesting concept. But does it take into account the whole twitter-sphere? If so, how does the site get all that data? Does Twitter pipe it to them or are they just following everyone on Twitter?
I have a domain name:
http://crowdsalon.com
better than crowdeye.com
may be you can use it.
We tried crowdsourcing for sales and It’ ’s not the best way!