A few weeks ago I had a chat with Robert Rebholz about folksonomies, RSS and information overload. It was a rather fun discussion and he let me know about a tool he'd built called the OPML-o-mater. The way the tool works is described below, 

The OPML-o-mater delivers a personalized list of RSS feeds in an xml format called OPML. OPML files can be imported by any competent RSS Reader/Aggregator.

You want the feeds from the OPML-o-mater because they're known quality feeds -- at least they were when we entered them. Even if you already have all the feeds you need, it might be worth a look to discover if we've one or two you didn't know about.

In general it works this way:

  • We've tagged the feeds.
  • You select the tags that describe your interests
  • The OPML-o-mater finds and displays feeds associated with the tags you've selected
  • You pick the feeds you want
  • Press the generate OPML button
  • Save the OPML file to your local machine
  • Import it into your feed reader
More specifically, we've tagged all the feeds. The first column of the OPML-o-mater lists the tags. You select a single tag from column one that describes an area of interest for you. Column two displays the tags that were also used anytime the tag you selected was used to describe a url (bear in mind that a single feed/url may have many tags associated with it). In column three the feeds associated with the selection made in column one are displayed.

I think this is a very interesting way to solve the "How Do I Find Interesting Blogs?" problem which plagues users of RSS readers today. I currently am subscribed to 158 feeds in RSS Bandit. Given that there are tens of millions of blogs out there, I am sure that there are literally thousands of blogs I'd love to read if I just knew about them. The tough question for me has been how to find them and then how to integrate that process into RSS Bandit in an automated way. 

What would be cool would be for the OPML-o-mater to evolve into the equivalent of http://del.icio.us/popular/ for RSS feeds and then for it to expose an API that tools such as RSS Bandit could integrate into part of their user experience. This idea is interesting enough that I wish I had the time and dedicated server resources to build it myself.  


 

Monday, January 30, 2006 11:36:57 AM (GMT Standard Time, UTC+00:00)
I am not really fond of the OPML approach since this adds another layer on top of the RSS concept.

What I would rather like to see is something like the following:

In my browser I want to tag a certain weblog, this would go into delicious or any other tagging community.

I want RSSBandit be able to subscribe to a aggregated feed of tags (something like http://yourtaggingspace/tags/web2.0+dotnet+programming+minkeys) and be able to auto-subscribe (with feed discovery!) to every new entry in this rss feed.

I do think that the traditional folder view of feeds won't really be of much use if taking such an approach. But since we basically allready have tags for the different rss feeds in RSSBandit (we retreived these from our online feed provider) we can perhaps have a 'tagcloud-like' approach at navigating your feeds.

What would even be better then this is having a feed dedicated to blog-posts that match certain criterea. But since structured blogging hasn't really flown off yet this is rather difficult to do right at this time.
Monday, January 30, 2006 5:04:39 PM (GMT Standard Time, UTC+00:00)
The idea of an OPML-based recommendation system seems interesting. But can't the major online aggregators (Bloglines, Google Reader, ...) already do this easily?

A dedicated service seems harder to put in place, mainly because of the bootstrap problem: how to generate good recommendations when you don't have many users. You'd need to crawl the web to collect as many OPMLs first.
Also, another problem is the use of multiple urls for almost the same feed of information (RSS vs. Atom, summary vs. full-text, ...), some feeds may even have unique tracking numbers in them...
Monday, January 30, 2006 5:21:56 PM (GMT Standard Time, UTC+00:00)
Julien,
Agreed. This is something that makes the most sense as a service provided by an existing online aggregator as opposed to a new service. However, none of the online aggregators has been particularly forward thinking in the functionality or APIs that they provide so I won't be holding my breath. :)
Monday, January 30, 2006 6:40:40 PM (GMT Standard Time, UTC+00:00)
If enough OPML is crawlable, then search engines could also create this functionality pretty easily.
But Bloglines OPMLs for example aren't crawlable. Bloglines publishes OPMLs, but they probably aren't linked for crawlers to pick up.

This is worth thinking a bit more about. I've been wanting to play with recommender systems ;-)
I was especially impressed with the simplicity of Daniel Lemire's recommendation system in PHP+SQL ( http://www.daniel-lemire.com/fr/abstracts/TRD01.html ). It does all the computation online, in real-time. But the computation may need to move offline to handle the scale issue.
Comments are closed.