Building a better Twitter directory

3

I saw (and participated) in a barrage of tweets today from Twitter directory WeFollow. Basically you send a tweet to @wefollow with #tag1 #tag2 #tag3 that you want to be listed under in the WeFollow directory. Then users looking for people who tweet about that topic can find the most popular tweetists by category.

I’ve kind of been watching the Twitter directory ideas for a while and so far I think this is the best attempt. As always, finding lists of the Twitterati (or anything else) works best when a person with domain knowledge and a comprehensive view of the choices picks the perfect list of whom to follow. Of course, these days no such person exists or has any way to comprehensively see anything, so we’ll have to stick to automated techniques that leverage human input instead.

Some things WeFollow is doing right:

  • No requirement to follow a magic user or give up credentials. Using an @ reply puts this info in the public tweet stream and makes it available for not just WeFollow but directories #2-100 to snarf the data.
  • Viral awareness – Using @ replies causes viral spread of the directory awareness. It rapidly will spread up to those who follow many people. And then it spreads down since those folks are commonly followed by many people as well. WeFollow has been up for 1 day and their top tag lists 3.6M people following. Clearly things have gone viral.
  • Open tagging – Any directory must support open tagging. Let tag customs and conventions emerge naturally.

And some things they (or whoever arrives next) will need to do better:

  • Difficult to distinguish between “interested in” vs “I am”. The #1 tag is “celebrity” (yawn). There are some actual celebrities on the list and naturally they are floating near the top. But it seems obvious that the 3.6M people “following” celebrity are in fact interested in following celebrities and are not themselves celebrities. The semantic web guys are sitting in the corner on a box of RDF weeping but I believe there must be cheap dumb ways to ferret out these basic kinds of distinctions without going all sem web.
  • No way to tag other people. The real way to build a directory that doesn’t suck is to take into account input not just from a user but from what other users tag someone. Note the relevance to previous point in fact. We are still at a pre-nascent stage for reputation markets but the rise of stuff like wefollow and SpeakerRate is a harbinger. Or you can look at the prominence of Twitter influence sites.
  • Not enough tags per person. I’m sorry, but three tags is not enough for me – I drastically over-tagged myself but I’m assuming that future directory bots will lap that up if wefollow doesn’t capitalize.
  • Forcing all tagging into the public stream is bad. Providing tagging input on an @ is good because the entry reqs are low (see point #1 on the first list). But it’s bad because it’s annoying to see every person do their tags. If you also want to push tagging for users other than yourself, that annoyance level is out of control.
  • Graduated tagging. It should be easy to say “I am X, Y, Z”. And harder to tag others. And harder still (but ever more valuable) to build your own niche lists.
  • Tag editing. While open tagging is essential to allow tag conventions to form, it’s also important to be able to retag later, merge tags, and get better at suggesting the best tag from possible options. Stack Overflow does probably the best job I’ve seen at pushing you towards better tagging by a) allowing question tags to be community editable and b) suggesting tags for new data c) with tag counts to indicate popularity.
  • Tag metadata. Intersect the directory with a tag wiki or comment stream to explain why you follow the tag, what the convention is, etc. Let tags accumulate links to explanations.
  • Tag networks – Tagging + network info = predictive communities. Once you have the two sets of info, you can start doing some very interesting data mining to automatically find groups and suggest tags to people that they may want to tag with or to find others that didn’t even choose a tag.

Follow me at @puredanger.

Comments

3 Responses to “Building a better Twitter directory”
  1. Eric Burke says:

    I don’t think putting the tags in the public stream is bad. That’s the main reason it is viral…people see everyone else replying to @wefollow, so they investigate and do it themselves. That’s how I found out about it.

  2. Alex says:

    I see it as both. It’s good from a viral and easy-to-start point of view. And bad if it’s creating a ton of @ traffic from people that’s annoying. At the current usage, I think it’s fine. But if it also contained tags for other people for example, then I think that would be too much.

  3. Nate says:

    Tags would be helpful in cutting down the chatter/banter that I despise so much from Twitter.

    I much prefer to use a decent News reader / RSS aggregator to follow topics that I’m really interested in, rather than follow the same bloggers on Twitter. Too much cruft.