Having been a power user of Bloglines for some time now, I’ve been thinking a great deal lately about personalization and recommendations engines in content aggregators. My information addiction has grown to the point that I can’t possibly manage the 300+ feeds manually added to Bloglines, and I need a system to help prioritize what I read.

The idea of intelligent newsfeeds has been present in fiction for some time now — custom-printed newspapers in J. Michael Straczynski’s Babylon 5 and preferenced newsfeeds in Warren Ellis‘ Transmetropolitan spring immediately to mind — but with today’s convenient common formats for sharing data (RSS), web technology is finally able to put this into wide practice.

Two of the early players in this market are Findory and Blogofy (Loomia is another of note, but that’s a topic for a later time).

Our personalization technology builds a homepage for each reader, recommending content based on what they’ve read and what new content is being published. We crawl through thousands of news and blogs articles so you don’t have to.

Findory’s key assumption is that clickthrough on a summary of content is a functional definition of interest in the complete content.

The Findory system bases its rating system on tracking user clicks, which is good for basic interest but isn’t necessarily accurate. For example, results could be skewed to the benefit of feeds which start out interesting, but turn out to be poorly written or simple rehashes of someone else’s content.

Worse is the frequent recommendation (and subsequent rewarding with higher ranking) of blog entries which are simple links saying “company a has released product b. i think it is cool.” Perhaps the preferencing of secondary sources is simply a function of how the blogging commmunity works, but I’d far prefer to be given interesting primary sources rather than simple reviews of them. With large amounts of data and users, these issues may work themselves out a bit better.

Findory allows “anonymous” (no log-in) preferencing, a very polite touch. I assume this is working through cookies, as it obviously isn’t truly anonymous, but providing this kind of guest access, Findory removes the login barrier to entry, allowing users to try this out without needing even to commit an email address.

Items specifically recommended by the personalization engine are noted with an icon which looks similar enough to the standard RSS feed icon standard to draw the eye and associate meaning, but is immediately distinguishable from that icon. It’s also integrated beautifully with the site logo.

I’ve been using Findory for a few weeks, and although I’m not entirely convinced it has learned much, it’s certainly a product which provides immediate satisfaction, holds interest with rapid change of its displayed articles, and provides interesting memes very quickly. It has currently joined Bloglines and GMail in the short list of pages I leave open all day.

While Findory has the look and feel of a reasonably mature product, I do have a long wishlist of things I’d like to see.

One UI flaw is the jarring page refresh at what is presumably a set interval; this often caused me to lose my place on a page, and could be made far more dynamic with a bit more javascript attention.

Pure text versions of rss feeds are also currently the only version of content available within findory itself, which forces the user to open links in new tabs/windows, where an inline version (framed?) of any feed providing full content might be much more convenient. This, however, is probably an intentional choice, although it creates problems with non-textual feeds (e.g., webcomics).

Findory currently lacks any “stop showing me articles like this” button, which would be very nice. If I could remove all sports feeds, I’d gain back a reasonable chunk of currently wasted real estate.

The current feed-vetting process is what I see as Findory’s largest problem, making it difficult to add feeds you already read for consideration in preferencing. Presumably, this is intended to be less of a problem as time goes by and more feeds have been added, but once the number of feeds reaches some critical mass, any further manual management will be problematic.

This editorial system also limits the sense of community given by fully user-driven success stories like del.icio.us, upcoming, and flickr. It should be easier to add feeds to their DB, as I have a lot of feeds currently in Bloglines I’d like Findory to track and recommend (and selfishly, I’d love Williamsburger to be recommended to other readers).

Blogofy

Traditional keyword based filtering engines unfortunately do not do a good job of weeding out the good from the bad. This prototype application goes a step ahead and uses a more advance algorithm to learn from user pattern to automatically filter news items based on user preference. The application allows each user to create his or her own account to train the system with.

Blogofy is based on the premise that your personal ratings of content can be used to determine what content you’d like to see in the future. This is a simple, good idea.

Unfortunately, the user interface is where this good idea falls apart entirely.

The ratings system is far too complicated for the end user. To help the user provide feedback, Blogofy displays a 0-100 rating next to each content item, and provides “more/less” buttons to raise or lower the perceived value. No user will have the ability to discern whether an article is better or worse than “13.65″ (yes, the 0-100 rating system goes into decimals).

With no mouseover explanations of the more/less links or the rating number, it’s difficult to determine what each element means. Worse, after some reading, it’s revealed that a lower value is better, but clicking “more” means you like the article and therefore want to lower the number.

Due in part to this incomprehensible rating system, there’s no sense of progress as you continue to rate links.

Additionally, articles don’t leave your “good” list once you’ve read and rated them, even after a page refresh, which means a lot of scrolling to find new things and no immediate satisfaction.

There’s also no ability to add new feeds yourself, leaving you limited to the ones they’ve decided are worthwhile. I’d assume this is just a feature that hasn’t been implemented yet, but as this kind of app should be targeting new adopters (who already have opinions about their favorite content sources), this strikes me as a must-have feature in order to build any sort of community.

Blogofy shares one of Findory’s issues: the user is forced to load content in a new tab/window rather than providing an inline viewer, even when the content is from a full-content RSS feed which could be displayed in an iframe.

In its defense, the site does state clearly that it’s in an alpha state; I would certainly agree with that, and argue that it shouldn’t have been made available to the general public at this stage. The idea behind Blogofy appears to have merit, but this version leaves much to be desired, as its UI is so obscure at this point that it’s hard to evaluate further.

Summary

Instinctively, I feel like I should like Blogofy’s ratings systems better; it’s a familiar, proven concept everywhere from Amazon to Netflix. Surprisingly, though, in its simplicity, Findory’s clickthrough ratings may be a better solution to a fairly difficult problem. Perhaps a merger of the two approaches could improve the accuracy of both measures?

This doesn’t get into the hairier issue of merging these concepts with social networks and trust levels, something the immensely intriguing but woefully buggy Outfoxed tries to accomplish, but that’s an entirely different discussion.

This entry was posted by Chris on Tuesday, January 17th, 2006 and is filed under Technology. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.