## War on Attention Poverty

Posted: 14 July 2010 in Uncategorized
Tags: , ,

Daniel Tunkelang has posted his slides from his talk at AT&T Labs on TunkRank over at the Noisy Channel. Embedded below for your viewing pleasure:

1. Ram says:

Hi Jason, I was wondering about some specifics in the TunkRank formula. For example how are p^{notice}, p^{retweet} computed. Since the computation is recursive, is the rank of a twitterer without any followers set to zero. Is this where the recursion ends? Details about such implementation specifics would be great. Thanks for your time.

I ignored $p^{notice}$ originally since it wasn’t a part of the algorithm when Daniel posted it. I continue to leave it out, since it can be factored out. If you assume a uniform probability for all users, then the scores are all scaled accordingly. That said, I have plans to estimate these probabilities rather than just assuming uniform probabilities. $p^{retweet} = 0.05$, which was an estimate I came to after discussing it with several members of the community. That one is probably a little lower, like maybe 2-3%, and again, I will be introducing better estimates in the future.

2. ram says:

hey thanks, Jason. I see you are still working on p^notice. Anyways, I see an update on tunkrank.com that shows top followers given a twitter id, that is those which contribute most attention. I was wondering about the algorithm/heuristic that is used to arrive at these twitter ids. I see these are different from most influential followers which can be easily elicited by comparing tunkranks of all followers.

Those are the followers who contribute the most to your tunkrank score. So it’s your followers ordered (descending) by $TunkRank(Y) / Following(Y)$.

4. [...] detailed explanation: http://thenoisychannel.com/2009/…Slide Version: http://mendicantbug.com/2010/07/…– TunkRank: We measure influence on Twitter based on how much attention your followers can [...]