Thinking and writing at the same time so I apologise if none of what I write makes any sense…
I’m pondering an alternative way of presenting Web feed subscriptions. Something different to the standard "There are N unread posts from feed foo". Are there any readers out there that will group unread posts based on their topic rather than on their source blog? I’m envisioning something that I can log into and see that "8 of your feeds have new entries talking about Google Calendar" (or whatever the big story of the moment happens to be). It wouldn’t necessarily be the best way to read all of your subscriptions, but it would be useful to get that sort of summary for anything that gets mentioned on more than, say, X% of your subscriptions in a short time. Like a sort of heads up on what’s the big news among all (or most, or many) of your subscriptions that you can instantly see before you start to read through all of the details.
As for the implementation, posts could be grouped if they contain links to the same URL; explicit meta-data tags or categories might be useful (either clustering all posts that share one or more tags or at least using them as a partial heuristic); or the system could use more sophisticated clustering techniques to try to figure out what posts are on the same topic. I think it’s reasonable to assume that a Web-based aggregator would be able to cluster posts very accurately given their access to a huge number of feeds. They could then present to each user just the subset of their results that relate to that user’s own subscriptions. Even for a desktop-based aggregator should have a good accuracy though, given the amount of data to base heuristics on (I mentioned tags and links, and even the plain text should give something useful).
Anyone else think that would be useful? Assuming a decent accuracy in grouping posts, are there any other features that could build on top of it? Finally, has this already been done and am I just making myself look dumb and out of touch by presenting it as new?