Two recurring themes have shown up in my development of RSS Bandit and usage of news aggregators in general

  1. There are feeds I'm susbscribed to whose content I never end up reading because there is too much content (e.g. Weblogs @ ASP.NET) thus missing the good stuff.

  2. There is no easy way to find content that I'd find interesting.

I've noticed more and more people complaining about the information overload that comes with being subscribed to too many feeds and wanting some way to sift through the information. I spoke to someone at work yesterday who said he'd stopped using his aggregator to subscribe to individual feeds but instead just subscribed to RSS feeds of search results on Feedster. Similarly some RSS Bandit users subscribe to a lot of feeds and just use Search Folders to sift through them. Both approaches are slight variations of the same thing. The first person would rather read all information in blogs about a certain topic or keyword while the other would like to read all information about a certain topic or keyword from a select list of feeds.

The goal of RSS Bandit is to encourage both approaches. For the former we provide functionality for viewing Feedster [and other search engines that return RSS feeds] search results in the same manner one would view an RSS feed. In the next version we will provide the functionality to directly subscribe to such search results in two clicks (type search term in address bar, click the search button, results come back as an RSS feed, click subscribe to search results). The last piece is currently missing from RSS Bandit but will be in the next version. For the latter scenario where users subscribe to lots of feeds but only read the ones that match the searches in particular search folders I am considering improving the search capabilities by supporting query-like functionality. Currently you can create a Search Folder that shows all items that match a particular key word or key phrase. However sometimes you want to perform searches over multiple terms (e.g. “Microsoft AND Longhorn”) or fine tune certain searches by ignoring posts that may coincidentally match your keyword but are not of interest (e.g. “Java -coffee -indonesia”).

As for finding new interesting content, RSS Bandit already provides a way to search for feeds by keyword on Syndic8 but there is a bunch more that can be done. There are a bunch of other ideas I have about enabling users to manage the deluge of feeds on the Web and finding new interesting content. Including

  1. Only show posts that have been linked to by other feeds you are subscribed to. This would work for news sites like Slashdot or high traffic feeds like Blogs @ MSDN.

  2. Add a way to integrate with Technorati's Interesting Blogs and Interesting Newcomers lists whenever they are implemented.  

  3. Only show posts that have a certain threshold of incoming links (e.g. 5 or more) as measured by Technorati. This may be infeasible due to causing high load on Technorati.

I'm supposed to be hanging out with Lili Cheng in the next couple of days, I wonder what she'll think of some of these ideas and perhaps she can set me on the right path.


 

Thursday, 04 March 2004 18:14:43 (GMT Standard Time, UTC+00:00)
We need some kind of a mix of tivo and slashdot for rss :)
Gia
Thursday, 04 March 2004 18:28:46 (GMT Standard Time, UTC+00:00)
This is part of the problem i commented on here. http://blogs.msdn.com/jledgard/archive/2004/02/11/71163.aspx

There needs to be more of a content/contributor rating attribution system so you can find the highly rated/commented/etc blogs.
Thursday, 04 March 2004 19:06:50 (GMT Standard Time, UTC+00:00)
1) Instead of using Technorati or other public blog ranking site, base popularity only on the content from the people I subscribe to. I subscribe to them because I trust and want to learn from *their* opinions.

2) Every time there's new .NET Rocks Show 20 people I subscribe to blog about it. I'd like to combine multiple blog entries about the same topic into one "logical" entry. This combined entry should have a higher rank than a topic only one person has talked about. How do you tell when multiple blog entries are about the same topic? That's Lili's job to figure out!

3) I subscribe to blogs.msdn.com AND to the blogs of some of the users who's posts are aggrigated there because I like to *scan* the titles from blogs.msdn.com but I like to *read* the entires from those individuals I subscribe to. Obviously there are duplicates, and those should be removed.
Thursday, 04 March 2004 19:38:14 (GMT Standard Time, UTC+00:00)
Dylan,

1.) How would this popularity manifest itself exactly? Do you want to be automatically subscribed to a blog if someone in your blogroll links to it consistently or do you want to use that to apply filters to blogs you are already subscribed to like I mentioned in my post.

2.) This sounds easy but is hard to implement in practice for a number of reasons. What does "merge" mean exactly? One uber entry with a merger of all the titles and all the content concatenated together? Or is this just some way of finding out what the hot topic folks in your blogroll are talking about without reading individual blogs?

3.) This isn't that hard to implement but I haven't figured out how to do so in performant manner. There is the added problem that blogs.msdn.com and weblogs.asp.net are synonyms for each other making it hard to identify posts in some cases.
Friday, 05 March 2004 03:06:34 (GMT Standard Time, UTC+00:00)
Is there a way to incorporate simple AI, for example Bayesian filter, and train it (by rating articles?)
A.Reader
Friday, 05 March 2004 04:35:18 (GMT Standard Time, UTC+00:00)
Hi Dare,

Thank you very much for this entry. I'll try and point to it tomorrow.

And we definitely appreciate the support in RSS Bandit. That's just great.

Best
Scott
Friday, 05 March 2004 18:19:52 (GMT Standard Time, UTC+00:00)
A.Reader,
I'm not exactly sure whether Bayesian filters would work as well for figuring out whether a post would be interesting to a users as well as they do for spam but it probably would be better than using keywords.
I'll take a look at bayesian filters when I start to work on this feature.
Sunday, 07 March 2004 05:57:28 (GMT Standard Time, UTC+00:00)
As was recently suggested elsewhere (I'm really sorry I have forgotten where), it would be nice to self-rate entries as they are read. This way, one could see which subscriptions rate highest to keep and which lowest to eliminate.

I want to weed out the noisy blogs, but don't know which those are at the end of the day.
Comments are closed.