FeedHub Launches - Individualized RSS Feeds
Personalization startup mSpoke is launching a new product to mashup and personalize RSS feeds today at DEMO [disclosure: Read/WriteTalk host Sean Ammirati works for mSpoke]. The product is called FeedHub and it creates an "individualized RSS feed" that aims to filter relevant posts from a set of feed sources. Like similar products we've profiled before on Read/WriteWeb - e.g. FeedBlendr, FeedRinse and BlastFeed - the result of the Feedhub process is a single RSS feed that you then add to your RSS Reader (e.g. Google Reader, Bloglines) or Start Page.
The reason behind FeedHub is to help users who can't keep up with all of the feeds they have subscribed to, by filtering for relevancy. Say you've subscribed to 100 feeds in Bloglines; by using FeedHub you can create a single feed that filters those 100 feeds. Ideally the resulting single feed will deliver you only the most relevant posts - and you can continue to 'train' FeedHub to refine this process.
My problem with such services in the past has been that the output, a single feed, is not very well integrated into a user's daily RSS reading experience. Ideally I'd like a service like FeedHub to be integrated into Bloglines or Google Reader itself (or whatever RSS Reader you use). In other words, a user ideally should be able to filter their feeds within their RSS Reader of choice. Perhaps we'll see that happen in the near future. However for now, you can subscribe to your FeedHub feed in any feed reader - including Google Reader, Bloglines and NetNewsWire. Another neat bit of functionality is that you can give FeedHub your clip/linkblog feed - for example your 'shared feed' from Google Reader - and have it learn from your interactions (see image below).
Memes
Feedhub is built on mSpoke’s "mPower Adaptive Personalization Engine", which the company has a patent pending on. The key to training a Feedhub feed is the concept of a 'meme' - popularized in the tech blogging world by news aggregator Techmeme. FeedHub will discover new memes for you and learn "meme weights" by noticing which posts you click on and interact with. You can manually adjust the weights associated with a meme. For an idea of how the algorithm works, here is mSpoke's explanation:
"Each meme represents some characteristic of a post - for example it's topic, popular tags in delicious and number of Diggs. Each meme also has a weight associated with it that indicates how predictive FeedHub expects it to be in choosing content you'll like."
How to use FeedHub
The first thing you do is upload your OPML file (or perhaps a selection of it). From that FeedHub builds a set of initial memes, based on your feeds. That gives you a single "individualized" feed such as this:
That's a lot of info, but you can also filter it down by adjusting your 'preferences'.
If we stopped there, FeedHub wouldn't be overly interesting - because adding that 'individualized feed' to your RSS Reader is (when it comes down to it) just another feed to track. However, it starts to get compelling in the next steps, when you interact with your feed from within the Feedhub website. Clicking on the 'Memes' tab displays this:
FeedHub monitors which posts within your personalized feed you read, and which you ignore. You can also provide negative feedback, by clicking on the links within your personalized feed to tell FeedHub "don't show items like this" or to "drop this source." Also by clicking and dragging memes, you can quickly and easily change your preferences. If you want more items from a given meme, you can drag the Meme to ‚ÄúYes, Please‚Ä?. Or if you‚Äôre tired of reading about something, you can drag that Meme to ‚ÄúNo Thanks‚Ä?. You can also add new memes manually, in addition to the ones FeedHub learns automatically from your behavior within the personalized feed. For example, you can tell FeedHub to recommend content to you that is popular on digg or delicous (e.g. the 'delicious hotlist'). FeedHub can also learn about the topics you personally tag in delicious or submit to Digg, if you choose to share that information on the digital identity discussed above.
Conclusion
We've said for some time now that filtering RSS is the next stage in the evolution of information processing. FeedHub is the latest attempt at this. It's not as integrated into my daily RSS reading experience as I'd like, but what you can do within the FeedHub website just about makes up for that. Also the concept of a meme set is compelling and goes beyond what I've seen other RSS filtering products do.
However try it out for yourself and let us know if it delivers the right results for you. In the final analysis, what counts isn't the methods of filtering - but whether it delivers relevant results.
Share2 TrackBacks
TrackBack URL for this entry: http://www.readwriteweb.com/cgi-bin/mt/mt-tb.cgi/1643
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
Seems like a good idea. I used to use Rojo for a similiar function. But it seems for the time being, I would rather trust the randomness of a feed ticker to go through all 500 of the feeds I follow than an algorithm to decide what is important to me.
Grereat service !
This one really helps to manage RSS easy.
I think this is going to be a good replacement for my current solution for my RSS needs.
Yours Sincerely.
Things start to get a little more interesting, once you start thinking about how you could mesh FeedHub with Yahoo! Pipes...
Stephan,
I recently joined mSpoke to work on the relevance side of the system.
If you're worried about missing data you like, you can request that FeedHub deliver all of your feed items. FeedHub will still learn what your interests are and report estimates of relevance using the "Spoke-o-meter." This threshold for delivery can be adjusted by you at any time.
The learning algorithm we use is transparent. You can view the "memes" (aspects of the feed item) that we use to estimate relevance. We also give you control to adjust the filtering algorithm. In the memes page for a personalized feed, you can also adjust the weights on the memes and create new memes to use for delivery.
We hope that transparency and control will build the trust of our users.
A warning to new readers, Feedhub is swamped. I just registered and Feedhub won't be able to put feeds through to me for 24 hours.
I'm hoping it lives up to it's promise of a learning algorithm. With the growth of the "social web" it seems there is less and less time each day to keep up with the various sites and services I have signed up with. If Feedhub can reduce the time I spend each day reading feeds by a reasonable percentage, I'll not only be happy with it, it will save Google Reader from a drastic pruning I had in mind.
Hi, Richard. Findory has had this since Feb 2006. You can create a single feed on Findory that combines all your feeds and recommends posts based on what you tend to read.
To use it, import OPML into Findory Favorites
http://findory.com/s/
and then select the "RSS Personalized" button at the bottom of the "Top Stories from My Favorites" page you get next.
These top stories are picked based on recency and your reading habits on Findory, so, as you read items on Findory or through this feed, the recommended items in the feed change.
Of course, just because Findory did it two years ago doesn't mean that FeedHub's version won't be successful. Findory's version may be cool and useful, but it is poorly known. It looks like FeedHub is being more successful on drawing attention to its version, so perhaps it will do better.
I've tried FeedHub. Yeah the first 24 hours are a wait, but after that it works. From what I can see you need to train it to show you what you like, like you do with last.fm as well. I imported the new feed into Google Reader. I think the idea is great. I think it will take a few weeks of perseverance before I know how good it really is. I reviewed it myself on my own blog yesterday but not as indept as here on Read/Write Web.