With the proliferation of memetrackers, I'm starting to wonder more about the algorithms that they use to determine the relationships between blog posts. Initially, I had assumed that they were purely link based: If A links to B, then A is about B. I think this is how TailRank works - so the differentiator that TailRank is after is mostly to do with coverage.
However, I don't think that this is how the Memeorandum suite of trackers works. Or, at least, they have this plus some other approach. The reason I believe this is that a post I wrote is currently listed as being in the discussion set associated with a post from Venture Beat about Powerset. However, my post didn't link to the Venture Beat post. So why is it in the discussion? In fact, my post isn't even about the Venture Beat post. The only real explanation is that Venture Beat linked to the Powerset website and so did I.
Unfortunately, I can't link to the article on TechMeme - permalinks there seem to provide a view of the discussion that is different from that given on the front page.
While it is interesting (and vital) that services like Memeorandum explore new approaches, cases like this bring in to focus issues of accuracy - how accurate is the description of the 'discussion' around a topic (which, here, means a post or news article)?