FriendFeed prediction - clustered feed data
March 18, 2008
Robert Scoble just switched his home pages from TechMeme to FriendFeed.
“So what?” is likely what you’re thinking. Yeah, big deal, right? Well, TechMeme had a clustering algorithm which would group together news articles of related content, and give you a good idea of the ‘hot topics’ of the day. It did this in a completely automated way.
I predict that FriendFeed (or another social network aggregator) will introduce topic clustering, based on the keywords and topics of people you follow. Clusty.com has done topic clustering for years, though it’s not something that is of great use to ‘general’ searching (at least, not in many cases). Carrot2, an open source clustering engine, also provides this sort of functionality.
I took a first stab at clustering my feed data with carrot2. I’m not sure I had enough data to draw useful conclusions yet - it might need a larger body of a group of people’s tweets (for example) which I just didn’t have at the time.
For people who follow thousands of users, it would obviously be useful to have a ‘big picture’ view of the hottest topics being twittered/blogged/etc about. But take it one step beyond that. Being able to look at *other peoples’* topic clusters would give you an instant view as to whether they have people worth following.
When I look at twitter, I can look at other people’s followers. Great concept, but it doesn’t tell me anything about the topics those people tend to twitter about, so I’m never sure if it’s worth following them. Nor do I get any notion of how those people are related. Marrying facebook or plaxo data against twitter feeds would be useful, no? Or just letting me add my own relationship metadata in to twitter itself.
Getting a high level view of peoples’ topiclusters would be incredibly useful. “Topiclusters” - yeah, I just made up that word and yeah, it’s lame. “Topsters”? “Substers?” (subject clusters?).
Did you like this post? Buy me a hot chocolate!
Posted in Blogging, Ideas, Search, Social, Software, Web, opensource
March 19th, 2008 at 5:18 am
Michael,
I can appreciate Scobleizer’s interest in trying to find something new through these aggregators. The unfortunate part about Memes created by some of the top read, followed tech personalities is that they create a remarkable amount of low-value noise after the first couple of coattail posts.
As for interfaces to quickly see these pearls, I think the RSS reader and river of news is great for following news updates, much like you watch the bottom ticker on CNN for quick bites. Problem here is there is an unbelievable amount of essentially duplicate items–clustering would be nice. This paradigm is also very slow compared to my ability to scan a Google News or TechMeme for interesting stories–the clustering is critical.
I like the popurls or new alltop.com paradigm for an interface, but again you only get the top noise makers, not who you may want to tune into. You idea of get a feed or page to tune into your favorite friends followed, clustered, content is great. Today, I am stuck with following or subscribing to all those trails as I see them and then cutting the ones I don’t like over time–very inefficient.
BTW, how have you been?
Bill
March 19th, 2008 at 5:32 am
Hey Bill! Doing OK and now down in Raleigh NC avoiding the snow (which my wife wants to get back to at some point)! Looks like things are going well for you - been following you now and then via blogs, and now via twitter.
I do think the clustered feeds will become a necessity at some point, but I’m not sure exactly what form it’ll take. It also seems like another one of those ideas that’ll just ‘click’ with people at some point and it’ll seem like a brilliant insight by whatever company adds it first (socialthing, friendfeed, google, yahoo, whoever).
Thanks for the feedback!