Building communities from Twitter posts

I spent a little time over the last couple of weeks playing around with some Twitter data. I was noticing how several people, myself included, were sharing the funny things their kids say sometimes:

So then I wondered whether there was a way to capture, prioritize and then syndicate the best Twitter posts into a ‘kiddie quote of the day’ or something like that.

My experiment only sort of works, but there are some lessons here that may be useful for community builders out there. Here’s what I did:

  1. Get the quotes: I ran some searches through Twitter Search and collected the RSS feeds from those results to create the pool of content to use for the project. In this case, I used ‘daughter said‘ and ‘son said‘. I put those feeds into Yahoo! Pipes and filtered out any posts with swear words. Then I had a basic RSS feed of quotes to work with.
  2. Prioritize the quotes: I’m not sure the best way to prioritize a collection of sources and content, but the group voting method may do what you want. Jon Udell has another approach for capturing trusted sources using For voting, there’s an open source Digg clone called Pligg. I set it up on a domain at Dreamhost (I called it KidTwits…Dreamhost has a one-click Pligg installer that works great) and then pumped the RSS feed I just made into it. In no time I had a view into all the Twitter posts which were wrapped in all the typical social media features I needed (voting, comments, RSS, bookmarking, etc.).
  3. Resyndicate the quotes to Twitter: While you might be able to draw people into the web site, it made more sense in this case to be present where interested people might be socializing already. First, I created a Twitter account called KidTwits. Then I took a feed from the web site and sent it through an auto-post utility called twitterfeed. Now the KidTwits Twitter account gets updated when new posts bubble up to the home page of
  4. Link everywhere possible: When building the feed into Pligg I made sure that the twitter ID of each post was captured. This then made it possible to “retweet” with their IDs intact. Thus, the source of the quote would then see the KidTwit posts in their Twitter replies. It works really well. People were showing up at the web site and replying to me on Twitter the same day I began the project.

    Again, I used Yahoo! Pipes to clean up and format the feed back out to Twitter to include the ‘RT’ and @userid prefix to each entry. I played around a bit before arriving at this format.

    I also included a Creative Commons copyright on all the pages of the web site to make sure the rights ownership issues were clear.

    Lastly, I added a search criteria for my feed collector that looks for references to KidTwits. This means people can post directly to the web site either by adding @kidtwits to their posts or #kt. There was already a New Zealand Twitter community forming who began using ‘kt’ to join their posts (short for kiwitweets), but they gave it up. I then had to filter out references to the kidtwits Twitter posts to avoid an infinite loop.

  5. Improve post quality: Now, here’s where things have been failing for me. I can’t think of better search terms to capture the pool of quotes I want, but there are so many extraneous Twitter posts using those words that it seems like I’m getting between 5% and 10% accuracy. Not bad, but certainly not good enough. The good news is that it’s pretty easy to kill the posts you don’t want through the Pligg interfaces. I just don’t have the time or desire to maintain that.
  6. Optimize the site: I then did a bunch of the little things that wrapped up the project. I added Google Analytics tracking, created a simple logo and favicon, customized the Twitter background, and configured Pligg to import the Twitter Search pipe automatically.

There are several things I like and a few I dislike about this little project.

  • I really like the fluidity of Twitter’s platform. It’s amazingly easy to capture and resyndicate Twitter posts.
  • I love the effects of the @reply mechanism. I can essentially notify anyone who gets their Twitter post listed on the home page of without lifting a finger. And they get credit automatically for their post.
  • I already knew this, but Yahoo! Pipes is just brilliant. I can’t imagine I would have even considered this project without it.
  • Pligg is pretty good, too. It does everything I want it to do.
  • I would love to hand over the management of the voting and quality checks to someone else. Voting naturally invites gaming. At the end of the day, however, the quality control and community management function is what makes a community service interesting to people. You can’t automate everything.
  • I’m actually not a fan of voting approaches to prioritizing content. It will ultimately result in dumbing down the quality. That’s less of an issue for highly niched topics like this, though.
  • The rights issues are a little weird. This wouldn’t be a problem in forming a community whose purpose is noncommercial naturally. But I’m not sure the Twitterverse would respond well to aggregators that make money off their posts without their knowledge or consent. (To be clear, KidTwits is not and never will be a commercial project…it’s just a fun experiment.)
  • Auto-retweeting feels a bit wrong. I wouldn’t be surprised if the KidTwits account gets banned. But I have explicitly included the source and clearly labeled each Twitter post with ‘RT’ to be clear about what I’m doing. I’m not building traffic to my account, the web site, nor am I intentionally misrepresenting anything.
  • By adding “RT @userid” I’ve killed the first 10 or so characters of the post that I’m retweeting. This means the punchline is often dropped which kills the meaning of the retweeted post.
  • Some conversational Twitter posts get through which include @replies to another user. When the KidTwits retweet of that post goes out it’s very confusing.

The potential here, among other things, is in creating cohesive topical communities around what people are saying on Twitter. You can easily imagine thousands of communities forming in similar ways around highly focused interest areas.

In this method the community doesn’t necessarily have the typical collective or person-to-person dynamics to it, but the core Twitter account can act as a facilitator of connections. It can actually create some of the authority dynamics people have been wanting to see. It becomes a broker of contextually relevant connections.

In a very similar way the web site serves as a service broker or activity driver. It’s a functional tool for filtering and fine-tuning the community experience at the edge. The web site is not a destination but more of a dashboard or a control panel for the network.

The experiment feels very unfinished to me still. There’s much more that can be done to create better activity brokering dynamics across the network through the combination of a Twitter account and a web site, I’m sure.