FriendFeed CommentRank

Comments are the essence of FriendFeed. For the partially completed crawl, I also started extracting the recent posts a user has commented on (or liked). All this information is available via the FriendFeed API. For a graph of about 22,000 users' *recent* comments network I ran the PageRank algorithm. Interestingly, the ranking is different again from the previous subscription only graph. I think CommentRank emphasizes more on the users who in the recent past have been active and surfacing cool posts that have attracted the community's interest. Hence it seems like CommentRank on FriendFeed might actually be quite fleeting! Here is a list of top 10 users according to CommentRank.

1. Igor The Troll
2. Mona N
3. Robert Scoble
4. edythe
5. Noah David Simon
6. Hao Chen
7. Shey
8. Tad Donaghe
9. Thomas Hawk
10. Mark Trapp

Again, these results are fresh of the grill and the crawl is still incomplete. As I tweak the algorithm and more crawl becomes available -- I guess, these rankings are bound to change. Although, Robert Scoble would be mighty glad to know that he has made it on both the lists! Wow! How does he manage to keep up with all that information?!

This gives me an idea to hack up a neat tool that shows the interesting FriendFeed users from the recent history. I think it would be a nice tool to help find new people to follow. Might actually do this if I find some time....

For now, this list shows the users as ranked purely based on the comment graph. However, I believe that the right way is to rank them is by combining the comment graph and the subscription graph. My reasoning is as follows:

"Influential/Interesting users on FriendFeed are those who can attract comments from people of importance/authority/high-pagerank in the network."

I think the nice thing about this dataset is that for the first time we can actually define as well as tangibly measure influence in social media as the ability of a user to attract attention (in the form of comments) not just by a lot of people but more importantly from authoritative users. This I call the InfluenceRank as opposed to the CommentRank (based on the comment graph) or the SubscriptionRank (based on the subscription graph). I think I would need to write some more code to get the Influential Users on FriendFeed i.e a modified version of PageRank that works on the comment graph but distributes the weights at each node based on the importance found using the subscription network.

This dataset is just fanscinating and exciting!

Comments

[Webomatica]

I use FriendFeed daily, and all those names are familiar to me as folks who highlight interesting stuff, very cool!

[Mark Krynsky]

This is some great Akshay. I think you're on to some exciting analysis of data that many of us FriendFeed users have been wanting. I really look forward to the crawl completion and an updates to the lists beyond 10 users.

[Akshay Java]

Webomatica, Thanks for confirming that these results are indeed useful. I was a bit concerned about that when I made the post. I use FF but I have still not familiarized myself with the subculture there so it was hard for me to know if this list meant anything. Mark, I am working on the crawl and also updating my algorithms. I hope to have some interesting follow up posts/demos using FF data. Stay tuned for more!

Thanks everyone for the encouraging comments on the two FF posts!

[liz]

if I haven't told u lately how much yr blog rocks: your blog rocks!!

I like this series of data you're mining on Friendfeed. Anxious to see what you come up with next...

[Akshay Java]

Hey liz, Thanks! I really appreciate your encouraging comments. I shall keep you posted as I gather more results. How are things with you?