I’ve just manually updated the Find That Feed database. 20,238 feeds plus there was a glitch importing one of the subscribers, so maybe more. Wow. Once this server has moved I’ll set about automating the updates, and figure out a carefully chosen new feature or two to implement. Btw, I’m looking for the best bang for the buck here, not to implement 100% (or even 50%) of all possible features. What can we do that SYO isn’t already, that would be easy to do with a relational database backend, and that would be, well, really cool?

11 thoughts on “”

  1. How about search results in RSS and OPML?

    How about, given two feeds, compute the likelihood that the appear together in someone’s feed-list? This is a rudimentary measure of “if you like one you’ll like the other”. Let’s call this measure affinity. If your feeds are x and y, call the sets of feed-lists containing them X and Y, respectively. Then

    affinity(x,y) = size(X intersection Y) / size(X union Y)

    Then if you cache this number for all pairs of feeds in the system, you can easily create a recommendations engine by suggesting the feeds with highest affinity to your existing feeds, excluding the ones you’re already subscribed to.

    I don’t know if that meets the ‘easy’ criterion; it is easy to compute a single “affinity” value, but doing this efficiently for all unordered pairs of feeds (20000^2 / 2) may be hard. Actually I think you should be able to do this proportional to X and Y in a relational DB…


  2. One I’ve looked for is “Most Unique Subscribers.” That is those subscribers who have the most unique reading lists. Something along an “Anti-Top 100 subscribers”.

    It should be easy to implement. (Easy for me to say, I’m not a programmer.) A subscriber’s “uniqueness number” = total number of users subscribed to all the subscribers listed feeds / number of subscribed feeds.

    OK. My math isn’t what it used to be. An example: If I subscribed to the following five feeds:

    1. elearnspace blog — 14 *
    2. A Thousand Words — 2 *
    3. Matt’s A.Whole — 10 *
    4. this is aaronland | weblog | rss — 3 *
    5. The Adventures of Accordion Guy in the –25

    (The last number on each line is total users subscribed to that feed.)

    My uniqueness number would be: (14+2+10+3+25)/5 or 10.8

    Of course, numbers closest to 1 would be the highest ranked. You could limit the list to the users subscribed to at least 50 feeds.

    Why would I like this list? It points me to the members most on their own. Their subscriptions are away from the mainstream or popular. I’d like to see why.


  3. Also, “bridge” or “gateway” feeds. If you can determine clusters of read feeds (these 7 feeds seem to be found together on subscription lists), there should be a few feeds that are common to readers of just a few of those clusters but not popular overall. They represent some idea, characteristic or personality that bridges the two clusters. Spokes connecting the clusters.


Comments are closed.