05 May 2004 | alan taylor said...

I like the way that Google News handles this - multiple distributors of essentially the same AP article get grouped together, with one representative source (maybe determined by popularity?) getting the lead, several variants getting "also-in" listings, and a link to the full set of nearly-identical listings. Just search Google News for " AP" to see good examples. I'd definitely be happy to see something like this. My newly-acquired addiction to Bloglines means that I see repetitive stuff constantly ;)

05 May 2004 | John H said...

Sounds like one of those obvious question/answer pairs which are never quite obvious enough, to me...

Logically the 'cheapest' way of achieving the Google-News-style grouping of posts would be to check related posts (and therefore their URIs, with the emphasis there on the 'U') against Blogdex, Technorati, et al. In fact, doesn't Technorati have a web services API for this?

I'd even code it, but I loathe news readers... the web is not Usenet, dammit! ;-)

[dons flame-retardent clothing]

05 May 2004 | Mark Fusco said...

So aren't you achieiving this when you use the subscribe to search function in readers such as Fastbuzz or Bloglines?

05 May 2004 | alan taylor said...

Mark, I've got a couple of subscribed Google News searches in Bloglines, and those things are the most redundant sources of news I see in a day.

05 May 2004 | Mark Fusco said...

Alan -

I guess what I am getting at is that the filtering is pretty much up to the user.

For instance, in Bloglines one can do a search for "x" and get a list of all the sites referencing or discussing "x". From that list of returns, one can then subscribe to the blog(s) they see best covers "x" topic from their perspective - and forget all the others.

Wouldn't that essentially be choosing a link destination?

05 May 2004 | One of several Steves said...

I guess it all depends on how you're using RSS. I use my reader to aggregate several of my regular news and commentary sources. I'm not sure if I have any blogs in there. If I do, it's only a couple, and I consult them only sporadically. So, that approach wouldn't work much for me. I can see how it might be helpful for people who use their readers as you do.

Perhaps more interesting (and much more complicated) would be topic groupings. That would cover circumstances such as what you mentioned, as well as people who use their readers as I do.

05 May 2004 | Lance Osborne said...

The adaptation of this feature in RSS readers would probably mean making changes to the RSS specifications, in theory perhaps adding a sub-element that points to an originating source.

("Specifications" is plural here, because there is more than one version and more than one "owner" of RSS.)

This, of course, opens one big can of worms, since it can normally take years for specifications to be reviewed, accepted, and then finally adopted.

05 May 2004 | Frederick Polk said...

...but it would be interesting to see a newsreader with an option to group entries based on link destination instead of link origination.

Well, how about you contact your client at Kinja and tell him that you would like to do a case study, with that feature added to the site.

Just a thought.

05 May 2004 | Brian said...

Another option would be to add something like a vector space search or a Bayesian categorizer to group similar items together. In this case, URLs would be just another "word" for the algorithms to grok. I haven't done much with Bayesian categorizers, but here's an example of a working example of a vector space search that categorizes posts on the fly. The downside is you need a rather large number of items before this sort of tool becomes handy. But dealing with a large number of items was the original problem anyways, right? ;)

05 May 2004 | Ste Grainer said...

Bloglines seems to be heading this direction. The other day, I noticed that some stories have "References" (with a number) that seem to act like externally tracked trackback links. I imagine with their centralized DB, they could easily do even more to tie linked entries together.

06 May 2004 | Frank Meeuwsen said...

In Newzcrawler (Newzcrawler.com) you can make watches. I think you can make them in Feeddemon too. It searches entries on keywords and aggregates them in a new channel. Not quite what you describe you want, but it can give you some pain of of clicking through a lot of channels...

07 May 2004 | Jack Vinson said...

You might want to look at SharpReader. Rick Klau mentioned it a while back: http://www.rklau.com/tins/archives/2004/04/16/sharpreader_threaded_rss.php.

It shows you which other articles in your aggregator reference the same articles. Not sure how it works, and I haven't used it myself.

07 May 2004 | One of several Steves said...

I use Sharp Reader. Its ability to aggregate similar articles is limited to articles from teh same source (e.g., only within the New York Times front page), and even then it's not consistent. I think it depends on how the source is tagging their stories. For instance, in my Wall Street Journal feed, multiple updates of the same story show up as separate stories. Whereas the NYT will usually cluster stories.

09 May 2004 | Alex said...

I've been using various RSS readers on both the Mac and the PC (yup, research). However I need some help with which one to recommend to others. I need one recommendation each of the following:
+ Mac OSX
+ Mac OS9
+ Windows 2000
+ Windows XP
+ Linux
+ Web Based (well, that's Kinja)

Thanks,

17 May 2004 | anu said...

Ahh - Feeddemon + the Group by style does exactly what you're after - group by destination links, showing how many posts are pointing at a particular link.

It's very cool.