Twitturly Cracks The TwitterMeme Nut
Michael Arrington
44 comments »
People who hang out on Twitter a lot know that quite often big news breaks there first. A recent example - when Chinese hackers took down SportsNetwork, the news was on Twitter well before we covered it.
But so far, unless you’re lucky enough to be following the right people, and online when the news breaks, you aren’t going to necessarily see the breaking news. Services like TwitLinks have launched recently that utterly failed to solve the problem, despite excitement from bloggers.
Other services, like TweetMeme and Quotably, are useful for tracking Twitter messages themselves. But the key is finding the useful links - Twitter messages are really too short to have much news value for the most part. And figuring out if two Twitter messages are actually related is very difficult, so the matching doesn’t work very well.
Today, though, Orli Yakuel pointed me to Twitturly, a service that holds some promise. It aggregates URLs linked in Twitter messages and puts them on the home page based on overall popularity, calculated simply by determining the number of times the URL was in a Twitter message. Like TechMeme, the more people who link to an item the higher it appears. As time goes on, the story deteriorates and drops in the rankings.
The result is a page of very fresh and interesting links that users can go to and see the most popular current URLs being linked to.
Of course what’s beautiful today is spam hell tomorrow. If this gets any traction (and I believe it will), it will have the same problems that Digg saw with people creating multiple accounts and linking to stuff just to bump up the votes. There are ways of dealing with this, such as giving more weight to Twitter accounts with a lot of followers, but it will be a constant battle against the bad guys.
Some of the results are also a bit questionable. One of the current headlines, for example, is to Twitter.com/login, which isn’t new or useful. My recommendation would be for the service to track URLs and only show “headlines” pointing to new stuff that hasn’t been shown in the service before.
We’ll see how it evolves. But for now, it’s a place to check out what’s interesting right now, according to the Twitter universe.
How is this different from Tweetmeme? Does TM track overall text where as Twitturly tracks only URLs?
Ben - exactly. tweetmeme (and quotably) track actual messages. Twitter messages are really too short to be interesting a lot of the time, and it’s hard to track connections between messages. Tracking actual links, though, is the way to go. That’s why I like the way Twitturly is going.
Twitturly : The way they present a comprehensive list of twitterers who posted the link is more useful for finding more people to follow. But I like the TwitLinks icon more!
Won’t this be taken advantage of by spammers pretty quickly?
It looks like the metric is active tweets, and not tweeters, mentioning the link. Does this mean one person can repeatedly tweet a link and it will get exposure? Could be trouble.
Seems pretty similar to Powncified (http://www.powncified.com), the major difference being that Twitter has enough traffic for the totals to be more significant.
I’m impressed that it deals with the popular use of tinyurls, snurls etc. on Twitter.
twitter is like digg to me but specialized to my own interest. I get the news I like and need. How do you think I found this post?
Looks great, but it also reveals the amount of twitter spam out there, and they will need to learn to filter this out.
Not only does their current homepage reveal spam for a 250GB hosting plan, but in the future, spammers can easily game Twitturly by opening up 100 twitter accounts and updating with their url.
Orli and Go2Web20 has become one of my top three sites now for the best reviews of new sites/services along with Techcrunch and Mashable. She consistently find things before anyone else does!
Glad you noticed how valuable she is.
Happy she’s come back to Twitter too!
Helped her test her account to ensure it was working right.
Welcome back, Orli!
Pai
http://memegator.com does this too (and has done since last year) - but it can’t be spammed as it only follows a select group of trusted tweeters
- you can’t submit anything to it - you can only see what the trusted tweeters are tweeting about - and it has lots of taggage too ….
ok it was developed over here in Limeyland by me, but I’ve found it very useful anyway
*dons flame-proof trousers*
The more video replies you leave, Mike, the more I see the value in them. Why? because when I have READ your comment replies on your blog in the past, and forgive me for this, i usually tend to just assume a snarky, sarcastic tone of voice (like most of what you say on the Gilmore Gang, lmao… [god it's nice to say Gilmore again])… usually people who get right to the point do so in an “eat this” fashion… and because when reading text the reader must assume the writers tone, it can quickly lead to being misunderstood and can eventually even lead to being disliked over time if people are constantly assuming an incorrect tone for someone else.
I wonder if anyone else has pre-fabricated “tones” that their inner monologue reads peoples comments as… I have them. I have them for Scoble, Arrington, Calacanis, Gray, etc… any common voice on the web I have a pre-determined tone of voice for that person… I know that’s wrong, but I can’t help it.
And, so as not to be 100% unrelated here, is it just me or could Twitt(url)y’s logo make a little more effort to not look similar to FriendFeed’s? (I know it’s different, side by side… but it invokes FriendFeed very much, no? I dunno… the logo I dislike.) I saw twitt(url)y last week and thought it was a pretty bomb ass approach… especially like the use of tinyurl/etc. APIs to resolve the URLs for comprehensive results.
So much for beautiful today:
http://twitturly.com/urlinfo/u.....cfbf32a33/
Re: Twitter SPAM, Can’t they just use the number of followers as a mitigating factor? Twitter spammers follow lots of people, but usually don’t get followed (unless they follow Arriginton or Scoble). Just as with Google’s “authority” ranking, twitter URLs could be weighted by the number of tweets, plus the number of recipients.
Wow… people just don’t watch the videos… lol.
Mike said in his video comment (#5) that following/followers metric would be a plausible benchmark for teaching their algorithm to detect SPAM…
Obviously just using “number of followers” is a poor metric. It would need to be a ratio of followers to followees (with a minimum). (and just displaying from a “white list” of select twitter users is both a cop-out and elitist and could be a major reason that the service mentioned in #11 has been around for a year and still has not gained global traction. People don’t want “select” they want to know that their vote counts, too. So the idea of a “trusted list” is hogwash. (Not a flame, no scorches … you can still get a full refund on those trousers.
)
Maybe they can hook up with TwerpScan to rate linkers..
I like it…. good idea. Definitely open for abuse though. I’d definitely say, weight based on followers should be taken into account.
Just for the record tweetmeme works exactly the same as Twitturl - we count the number of references to the same url - but the way we style it is very different to them.
But we also categorize the content ‘blog’,'video’,'image’ etc,
A big update soon as well with spam filtering / better relevancy / language choices + a load more.
Just have to finish this other little project called fav.or.it at the same time….
i agree with the comments regarding spam. with picobuzz.com i am concerned about how to block the growing spam or autobot twits. Sometimes the volume of buzz on hot topics compensates, though, and drowns out the spammers.
@#14 .. in re-reading your comment I drew a new meaning… and i like what your saying a lot…
The only thing to watch out for is the Scoble effect… he twitters a link and all of a sudden it’s number one because 70% of twitter users got the link at once. lol.
Another thing is the Calacanis effect… he links almost strictly to Mahalo, so the system would be uber-cool if the system would substract rank based on the number of links a given user tweets to a specific domain.
Example: Scoble links to 150 different sites today. Each one gets full on Scobleized. Calacanis links to Mahalo 713 times today… but each consecutive link gets “knocked down” a peg lower than the last one, because it links to the same domain (not URL… domain). By the end of his linking frenzied day his 750th link is all but ignored by the system… i think that would provide a nice counter measure to the “added rank based on # of followers” … then if any Mahalo link gets re-tweeted by a ton of people then it will find it’s way to the top, it just won’t get there automatically based on the popularity of the person issuing the tweet.
@#18 .. wow, i just visited tweetmeme homepage and I gotta say… I’m not convinced there is any algorithm at work there at all… what a disorganized mess. Furthermore, they strip the URLs out of the tweets… that’s bad form as far as I’m concerned. So no, it’s not the same thing at all. Not to mention, it doesn’t even work. lmao. Twitt(url)y works… tweetmeme looks like it TRIED to be pretty and THEN decided to throw a little functionality at the thing (and missed).
I just found twerpscan on twitturly and used it to block 52 obvious spammers on my twitter account. Thanks Mike, twerpscan and twitturly!
who. cares.
http://www.penny-arcade.com/im.....080423.jpg
Forgot to mention -
We also have a river of news -
http://www.tweetmeme.com/river.php
or a feed to subscribe to - http://www.tweetmeme.com/rss/all.xml
And you can (like techmeme) review what the site looked like on any date.
RSS…Please keep a Feed for this service!!
Oh man, this feels like an echo chamber inside an echo chamber…
I created a feed for it using Dapper:
http://tinyurl.com/6yh9tz
@#24: Nick
yup… even with those links tweetmeme still reeks like wet dog ass. not to mention, if your list is not providing useful results on today why would anyone care what it looked like on a different day? Mike posted about twitt(url)y not tweetmeme (be glad the service was even linked to in the post… a simple “Thank You for the link!” should have sufficed)… you’re really coming off as a spammer. (Come to think of it, you spam the whole web with your shit, too… when someone posts about another service, just let it be… don’t rush into every single blog post and spam everyone… you do it all the time and it’s annoying. god help the poor people who work on your team… they must hate your routine as it makes them (vis-a-vis your products) look like whiny complainers who can’t stomach a little competition.) So god, just realx a little and let some other people play in “your” sandbox, too. You big meany.
hehehe. (Just kidding about the meany part, of course. The spammer part? Not kidding so much.)
@28. Matt, point taken - I feel justified in correcting the mistake that the services in work the same way, but talking any further about what we do was wrong and I complain when people do it on stories about us, so slapped wrist..
Thought I’d mention a site I happened to launch the same day Twitturly did.
http://intwition.com/
You can do a little more filtering based off ‘Reach’ (which is a cumulative sum of how many timelines the post reached - or total number of followers for all that posted).
The original focus of the site was to allow people to easily track what was being said about sites on twitter by providing RSS feeds of comments and stats about each site. But as I started sorting through the total data I saw that it could easily be used to show popular items per day.
Obviously battling spam has been hard, but by filtering using ‘Reach’ you can drop off a lot of it right out of the gate.
@#29 .. damn… such a civil reply I almost feel like apologizing. (…almost
).
@Matt
You sound like a whiney bitch yourself. And I’ve never seen you comment before today (the day of twitturly). I would guess you have a direct connection considering your claimed knowledge. I dont like either service, both of you STFU.
Maybe they should have called it twigg
@Michael Arrington First, thank you for the write up. Second, I intend on adding the freshness capability that you mentioned in the article and your video as well as weighing in the number of followers.
I already have your weighting suggestion built in, but for now it is turned off because the spam system is learning the characteristics of spam messages. Although overall the spam system is still quite rudimentary, there are quite a few systems that I have built into Twitturly prevent spamming of the system. I have user blacklists, site blacklists, comment interpretation, discounting duplicate comments from multiple people, single tweet counting from the same user and I recently added bayesian learning (although the bayesian filter is off while learning). I also have a few more things that I will be releasing soon dealing with spam, but I will not say so that people don’t know what they are.
I am not sure yet what a good freshness setting would be, whether it is just 7 days or 30 days or something else. But I will play with it and figure out a good setting so that Twitturly always has a good list of the most recently talked about.
@Jeff Doak and @kuldeep There are already RSS feeds on Twitturly. You did not need to create a dapper feed, but thanks anyways.
@Nick Halstead Change the language filtering to “Unknown” and you can see just the images, audio, movies, etc.
@Matt (#20) I will look into how this can be done and if feasible, maybe I will do it.
The story you linked to about Twitlinks said “It may even be the only news aggregator that is faster than Techmeme”.. Well just today in the late afternoon I found a very useful story via Twitturly..only to see it come up several hours later on TechMeme.. Maybe it’s _Twitturly_ that’s the real deal. Twitturly out-memes TechMeme for Twitter? Too many T’s.
Ooo another target for social media optimization. I really love the idea and premise though - I wonder what other applied APIs we will see to the Twitter arena =oD Fun stuff and actually a great tool for publishers of content - Good news stories likely come from Twitter users