The answer is not a better RSS Aggregator, continued
Yesterday's post drew some interesting comments and emails. Several people wondered what the heck I was talking about. More about that later. First:
A couple of good analogies were raised. Isn't this idea a lot like Digg? Or like Top Ten Sources? For me the Top Ten Sources, which I know reasonably well, is the better analogy. Certainly Digg is hot but the reason it doesn't work for what I am talking about is that it bring you links to content or sites that (random) others have marked as interesting.
A key part of the idea that I am groping for is that the user somehow tells what he or she is interested in.
Example: Alice is interested in:
- Cars
- Any article that Tom Friedman writes
- News about Belgium
- Classical Music
And with that very minimal bit of guidance this new product or service, if it is ever built, starts delivering articles, pictures, information to alice, Alice. And this info includes sources that she never heard of before nor had any way to locate or read. A little bit like she opens her weekly Time Magazine or Car and Driver every month.
So yeah, Top Ten Sources is a little bit like that isn't it? Maybe. Maybe not.
[Let me stress again, these are literally musings. We are not building this product. But we might.]
Technorati Tags: feed.tv, dailyme, readinglists
Comments
Here are some thoughts: Why should Alice have to enter that list of interests anyway? What if Alice's interests change -- is she responsible for maintaining that list of interests? And what if Alice would be hugely interested in something that she doesn't even know about -- how will the system help her to discover it?
How about an attention-based model, where the system determines things you'd be interested in by correlating your attention data (which stories you actually read/click) with that of other users of the system. If you read a lot of Star Trek news, and the system knows that most Star Trek fans also like Battlestar Galactica, the system might send some Battlestar Galactica news your way. Think last.fm, only for news/blogs/etc.
Posted by: Kevin Yank | October 24, 2006 12:24 AM
This is a FASCINATING discussion (as I Google Bookmark and add to Reader...both of which blow in terms of context-sensitivity)...Have you reviewed the Haystack project (http://haystack.lcs.mit.edu and http://haystack.csail.mit.edu/overview.html) (based on Eclipse)? RSS/RDF/whatever...all in one place. How is this relevant to RSS aggregation? Well...
From the _context_ of each item, their proximity and "time-spent-on" to others, one might begin (using NLP and social, Web 2.0 "friends" contexts) to not only deliver the same content, but the same _contexts_. This might also yield interesting results when allowed to "think outside the box." Think an "inverse SPAM filter"...if most SPAM has a relatively high noise-to-signal ratio, the random suggestion in a relevant context made by a friend via his/her blog/e-mail/search history might provide an item from amongst 1000's of feeds you have never read because of the "oh, by the way, I saw this the other day...cool, huh?"
It's awfully late to be more coherent, but I'll be watching this one closely. Blog on, good soldier!
-j
Posted by: Jeff Hampton | November 14, 2006 02:22 AM