3+ year old TechMeme, an automated news site that shows breaking news clustered by topic, has always generated “headlines” by analyzing how news sites link to each other. If a lot of sites start linking to something unique, TechMeme guesses it’s news.
That isn’t working, says founder Gabe Rivera today in a blog post: “Only an algorithm would feature news about Anna Nicole Smith’s hospitalization after she’s already been declared dead, as our automated celeb news site WeSmirch did last year.”
He’s hired someone to start vetting stories that the algorithm says are headlines, to either push them up or get them off the site entirely.
I believe this is a slippery slope for TechMeme. Certainly a human editor can make the results better. But it also completely destroys the objective nature of TechMeme and turns it into something different. It’s now subjective, and in many ways just another news site.
Today the TechMeme Leaderboard (and we track individual authors on CrunchBase) is an objective list of what sites are breaking the most news and getting other trusted news sources (”trusted” being defined by TechMeme) to link to them. With these changes it isn’t clear what authority that Leaderboard commands.
Overall I think we can expect better and faster news on TechMeme, which is an invaluable resource for journalists, bloggers and news junkies. But the fundamental nature of it has changed.
A little manual help never hurts…
That’s what she said.
Make it a full community ala digg and you get manual help for free.
Not surprising. Prioritizing ‘news’ just by the number of links is a very one dimensional perspective. Otherwise you could just use goof search.
It all depends on how involved in the process this administrator gets. If all he does is remove faulty stuff then I see no problem. If he is constantly moving stuff around then you have a point.
I think a lot of people would like to hear more about the algorithms that drive Techmeme by the way.
Funny, I just caught this story from another automated tech news aggregator – http://technewstube.com/
That is the uglist list of headlines I have seen in quite a while. I didn’t even stay long enough to check if its a mere river of news instead of an actual aggregator.
Well, I gave up in 2005. Just announced it today.
I’ve always been ambivalent about the Leaderboard myself. See for instance my post here: bit.ly/YBg0
Looking below, it doesn’t seem anyone read this comment.
So, you are saying that Arrington’s complaint is “retroactive”, and would have been valid for the last 3 years?
Next thing you know Techmeme will start asking readers to vote headlines up or down
And comment on them, too.
Does not this open TechMeme ot more gaming of the system?
Whoever this new admin guy is, he’s going to get a lot of lunch offers and party invites from now on.
It’s a she.. Megan McCarthy, formerly of Wired.
Wasn’t she the Party Girl of the well-respected and much-loved ValleyWag?
My understanding is that he is a she; but yes, she’ll likely be getting a lot of attention.
Judging the QUALITY of the coverage by the QUANTITY of the links is a generally flawed idea. Page was right when he introduced it for the general search. He viewed the Internet as a ‘library’ and in general CI was a well forgotten old concept, he just applied it to the pile of junk instead of the meaninful scientific articles (which undergo editorial approval before they are published).
As to the news, it would work too if all the sources would be: a.) republishing the same; b.) dilligently linking. They are not. They are so deep in the s…t of posting meaningless links (that’s how they do ‘the SEO’) that no algorithm will work. Also, if it would people would invent a workaround immediately. It’s the same thing as with a ‘automated trading’ on stock exchanges. Algorithm CAN work for a brief period of time then, when everybody starts taking it into account it just stops working. Same thing with all these ‘fully automated’ bla-bla-bla.
The mass media does something great: it filters and aggregates.
So people thought a computer could fulfil the function. And to date, every startup I know that has attempted some form of this has failed. I don’t wish to bring up any names because I am good friends with a half dozen of these companies, but if you sit down and think about it, it’s pretty clear.
That is, computers still can’t replace humans (and the closest we will ever get to that is when the semantic web comes to life in a few decades). I think the announcement with Gabe is consistent with what I think is reality.
I was thinking the same thing last night for some reason – that the only good recommendation engine you have to this day is people passing along news, information and links to each other. Even Digg cannot begin to approach the quality of information that is found by one person passing along information to another. Any attempt to automate the process or assign algorithms to it still screws up everything – the only thing automated aggregation helps with is spreading the word about something worthwhile *further* – it cannot discern how worthwhile something is, though.
Interesting – I think this really does change the fundamental premise of Techmeme. Objectivity is gone and replaced by a single editor. Say what you will about other social news sites, but a diverse crowd is definately more democratic than a single editor (sorry Megan).
There are sites that are trying to define quality/credibility, like NewsCred and NewsTrust but there are separate challenges there.
Very good move.
trying to see if facebook connect works
Gabe:
You (and Megan) are welcome to come on over to socialmedian where we have 2382 entirely-user-edited techmemes on topics from web 2.0 to obama to video production to interactive marketing, etc.
It’s a nice scalable model.
http://www.soci...twork?s=popular
With kind regards from one of your greatest admirers,
-jason
Oh wow. same comment using FB Connect.
to prove it’s really me.
You (and Megan) are welcome to come on over to socialmedian where we have 2382 entirely-user-edited techmemes on topics from web 2.0 to obama to video production to interactive marketing, etc.
It’s a nice scalable model.
http://www.soci...twork?s=popular
With kind regards from one of your greatest admirers,
-jason
Hmmmm
I’m cribbing my own comment from AVC last month:
Here’s something for you: I run a memetracker as well, its a weekend project for me that I’ve been fooling around with for a year or two. About a month ago, I started using the Twitter API to post items from my tracker to twitter, and was pleased to see some take up.
Now, though, I’ve taken it a step further: Twitter provides attention data that feeds back into my algorithm. Because each story is presented identically on Twitter, I view twitter click-throughs as great organic indicators of interest – i.e.: clicks aren’t driven by page placement, font size or what have you. The twitter attention data seems to be driving some interesting results, generating a page that covers many of the same things as Techmeme, but that also surfaces some really different things.
Have a look at the feeds: http://twitter....om/techwatching and http://twitter....techwatching_cl
kool
I wonder is he was using RSS 2 Blog for his automated blogging…lol
Unless an all knowing being created the algorithm than it is still subjective, just consistently so. Thankfully humans still have a place in the world.
test
I like the idea of including a human editor, partially because I don’t think that a purely algorithmic approach can truly be objective in the first place. The creation of an algorithm includes decisions that create inherent bias, and usually encourages some sort of gaming of the system.
I personally think the model of using a human+algorithm combo is a really a great approach to news delivery. I’ve been working on a similar effort focused on aggregating sports content (buzztap.com) which will also soon be adding a human editor. If you’re a sports fan, check it out.
Do you guys really still actively track all authors on CrunchBase? I figured it stopped updating a while back because my numbers and the numbers of the few others I’ve checked were way off.
We do the same thing like Techmeme.com. We use human editor and automatic program to collect news. We think it is a good combination to sort our the best news as long as the human editor is not biased.
i think that it’s inevitably algorithms plus humans — it’s the best way to the highest quality product.
I agree. But it’s also subjective by nature, which means it’s not, well, objective.
So TC is objective?
TechCrunch Facebook commenting = awesome.
Techmeme Automation = awesomely funny.
http://www.flic...ell/2988532013/
I wish Gabe and Megan luck with their hybridized cyborg solution.
If the human input is focused on removing clearly bad results rather than exercising wide editorial powers, then it should be fine.
Can look at it as a human overseeing a machine in a factory. If the machine messes up every once in a while and makes a bad widget – you need to pull that widget out and trash it. Basic QC.
No such thing is white and black. Practicality >> ideology.
What you missed is that Gabe hired Megan McCarthy, formerly of Valleywag.
I wonder if the editor will, for instance, ensure Techmeme spotlights a broader slice of the news pie, and not just headline after headline from the same six sources.
“Objective” and “subjective” — you got it backwards, Michael.
Gabe was critiquing what should be published as “news”, and he was describing which practices got him closer to that goal.
Gabe suggested that the algorithm (et al) didn’t measure “news” very well, and that involving an editor did better. If we believe this, then this actually makes the human hybrid system more objective, not less. There is more likelihood that multiple people will agree that the published stories are indeed “news”, making them more like external ‘objects’ of the “news” variety, than in the algorithm scenario, which produced an unreliable and idiosyncratic notion of which stories were “news” based on its own idiosyncratic machinations (i.e. cross-postings, google, etc., making it more “subjective”).
I don’t think that an algorithm can be “objective”, it’ll always follow subjective rules that were assigned by its creator.
I do believe though that the solution to this whole “automated news” enigma, is a mix between several “filters”. The algorithm is a filter, an editor is another filter, digg-like systems are also another kind of filter etc…
Next thing you know Techmeme will win a Crunchie for the fastest growing startup in terms of percentage. Last year they won the best bootstrapped startup with 1 employee and now with their explosive growth they are up to 3.
Massive growth
I think they should continue to work the algorithm. It can work. There is another way keep trying Gabe.
Michael,
For fully automated news check this one:
http://www.news...pics/world.html
- the hot topic cloud gets auto-generated every few minutes
The threading model for Techmeme story links is a little wonky…
A little primer would help.
There are the “original” news sources and then there are the me-too-blog-wankers… how does one provide merit for the originator of the news?
TechMeMe is great tools good for tech savvy.
keep it automatic, that´s my vote
I would totally support using human editors to make a few judgements about the validity of the news.
Every news source republishes (updates) news. A machine would not know if its updated or real breaking news…