Comments for Track Hot Topics On Niche Blogs With FeedVis

Track Hot Topics On Niche Blogs With FeedVis

2009-01-19T23:51:29Z

Want to put your ear to the ground and find out what any group of bloggers are talking about? Some types of bloggers link out to each other a lot, making it easy to see what the hot topics are (see Techmeme, or Technorati). In some circles, though, blogs don't link to other blogs' posts regularly. That's why there will only be so many variations of Techmeme and why we need other tools if we want to track conversations in other parts of the blogging world.

That's what a new service called FeedVis offers. Give it a bundle of blogs (in OPML format) and it will give you a scrollable, searchable tag cloud - a visual representation of the most-used words in a given period of time among a defined group of blogs on any topic.

]]> Put it on Your Server

Users can put FeedVis on their servers, upload a collection of blogs in OPML format and add a list of "stop words" to exclude from those terms from the popularity contest. The end result is a beautiful scrollable and searchable chart of the most popular words being used in one day, a seven day period or throughout the recent history of that collection of blogs. Increases and decreases in usage are displayed and clicking on any word will bring up the recent blog posts where it was found.

There's a demo site available at FeedVis for up to 100 blogs, but there's not a way to upload a list of words to exclude so the analysis is of limited quality. In the example above, for example, you can see some very likely candidates for words that should be excluded from a group of education bloggers. You can see the interface on the demonstration site though and it looks quite handy.

The source code to put FeedVis on your own server is quite simple and easy to customize. We think a lot of people could find this service useful.

Comment from Wyatt on 2009-01-19

2009-01-20T01:27:05Z

Looks like a handy service, I could see myself using it, though the demonstration suggests it might take a bit too much customization to be useful - probably seventy percent of the keywords suggested aren't actual trends, but common words.

I think the system may need to incorporate a bit of intelligence before it becomes truly useful.

Comment from Marshall Kirkpatrick on 2009-01-19

2009-01-20T01:30:56Z

Wyatt, that was my first impression, but it looks quite simple to train when it's on your own server. Just knock out the most common words in your niche and the standard of what words are most commonly used will change.

Comment from Engago Team on 2009-01-20

2009-01-20T12:23:50Z

Why does FeedVis require an OPML and not just use the blog feed?

Comment from Martin B. on 2009-01-20

2009-01-20T21:25:04Z

Hi,

thanks for the introduction to FeedVis. The project sounds really interesting and, it has to be said, really fresh, given that everyone is in a total tag-frenzy nowadays.

I have to agree I'm not quite sure if it's intricate enough to deal with such complex stuff as texts, words and topics. It seems to give very meaningful words, but there is little information about _how_ they're used (i.e. just mentioned once/post.. or many times per post), which could be interesting too.

Following up these thoughts in my blog, so click on my name.. (including a trackback to you) :)

Cheers,
Martin

Comment from Jason Priem on 2009-01-20

2009-01-20T21:36:56Z

Hi, it's Jason, the guy who wrote FeedVis. Thanks for the post; I'm glad you got something out of my project. FeedVis is still very much beta software, so any suggestions anyone has are always appreciated.

@Engago Team: FeedVis is really more for communities than for a single blog. If you're keen to track just one feed, though, you can do it by making an opml for just that blog; several feedvis.com users have done just that.

Comment from Jason Priem on 2009-01-20

2009-01-20T22:13:12Z

@Martin B: Ultimately, I think we're going to end up with very, very complicated algorithms and AI to do metadata generation; it's just a really hard problem. In the mean time, we're stuck with relatively clumsy statistical techniques. However, I agree that "just countin' frequencies" is clumsy even for clumsy. I'd like to engage the texts on a deeper level.

I don't have a lot of time to work on FeedVis lately; it's just a for-fun, spare time kind of project. More sophisticated text-mining is definitely a to-do for the next version, though.

Comment from Martin B. on 2009-01-21

2009-01-21T18:09:49Z

@Jason: Totally agree, you say it's "clumsy even for clumsy", but the tool is still very interesting and obviously the results make sense. So I'm not so sure how complicated the algorithms really have to become here, we rather need inventive ideas and bold experiments to get us further in my opinion.

Keep up the good work! :)