There’s an ongoing conversation on the Gillmor Gang around Twitter’s disabled track functionality. This functionality allowed you to set a search term that would be applied to all messages (tweets) sent in Twitter. Matching tweets would be sent to you via the Google Talk interface. The user experience was that you could open up your IM client, tell Twitter that you want to get all messages that contain “core php” and they would appear in a chat window in real time.
Unfortunately, this functionality apparently wasn’t coded to scale with the popularity it attracted. Twitter took the feature away months ago and it hasn’t returned. You can get a similar experience by subscribing to a search in Summize, a search engine for tweets. In this model, you enter a search term and then subscribe to the results via RSS. Your RSS client can poll the feed regularly. But the experience is much slower. I understand that while track was working, Steve Gillmor would watch tweets flow in and carry on realtime conversations. This isn’t practical with saved searches. You can check your feed once a day or so. It’s a half solution.
The real weakness of both track and Summize is that they take an old school way of finding information. Remember when Web page search engines just indexed the content? It worked for a little while, then people optimized content to get undue attention. That is, they put a bunch of keywords on pages and pushed their results to the top. This wasn’t helpful to people looking for information.
A better search for tweets would involve gestures by listeners. Perhaps an impartial observer could contribute as well. Additionally, search is always behind. It’s a report in retrospect. The delay was short with track, but what we really want is a system that suggests information we might be interested in. That will allows us to learn things we don’t yet know we want to learn.
Here’s my vision. When I follow someone on twitter, I should also answer the question of why. I would do this with tags. When I subscribe to a feed in Google Reader, I put the feed into a folder that answers why I subscribed. Egg City Radio is in my “music” folder. Techcrunch is in my “vendor-sports” folder. With some minimal set of people, the gestures would add up to an aggregated rating of content topics.
I would expect that the system would also benefit from an analysis of the content. Someone who talks frequently on a topic should be tagged by an automatic process. This would fill in gaps before people discovered the speaker. Gestures by people would override any tags generated by content.
The user experience I’m looking for is to ask Twitter for people who talk about a topic I’m interested in, liberated bootlegs for example. Twitter would suggest 10 people I should follow. Google does this with Reader. My universe of topics I pay attention to in Reader probably would not match completely with the universe of topics I want in Twitter. I imagine I’d want more local news and politics in Twitter and less about big chunks of entertainment.
I don’t see what would prevent this system I describe except Twitter continuing to control the attention of microbloggers and refusing to implement it. Identi.ca certainly can implement this feature, but they better match Twitter’s features first. I suppose Google could ramp up Jaiku to include it, and it might be easy given it exists in Reader.
. Programming techniques are one thing, but it’s harder to acquire an understanding for history and lore of the craft.