In some ways, Twitter search is the blasphemous opposite of Google search.
Google analyzes the number of incoming links to a source, and the number of incoming links to those incoming links, and estimates authority based on that link capital. When someone searches for something on Google, Google ranking URLs higher on the results page if they have more links from more authoritative sources.
Twitter ignored all of this art and science in their real-time search, and still managed to create an insanely useful resource. In its version of search, Twitter pulls the most recent content items, and presents them in reverse chronological order, based solely on exactly matching keywords.
It works. But it could work better.
Take, for example, the #IranElection- of #Haiti-related explosions of content on Twitter. We become instantly flooded by real-time content, and ostensibly relevant updates, but of completely variable, unverified authority, to the extent that the firehose becomes useless. The signal-to-noise ratio is nil.
The Premise
Twitter’s greatest asset is its selective, curatorial structure. I chose people who know what’s going on, like AmanpourCNN and BrianStelter. And they are following people who know what’s going on. The result is that my personal feed on Twitter has high quality, breaking information that tells me more about what’s going on in 30 seconds than 30 minutes on a news site, or an hour reading a newspaper, could do.
But when I want to search for something, the scenario changes. It’s the messy Wild West, a clamoring thrall of repetitive messages, some spam, some false, from every direction.
Searching only my followers would help, but it’s not the solution. True search will let me look beyond my curated network to tell me who the best experts on Twitter are, and what they say. The good thing is, all this data already exists.
The Proposal
If Twitter’s greatest asset is its user curation of insightful, informed people, why not leverage this in search?
Users as Sources and Followers of Followers
When I search, show me the most relevant, real-time tweets from experts. And filter those experts, or sources, based on their influence. This means they should be filtered not only by the number of followers, by the number of followers of followers — the equivalent to Google’s ‘links to links’ – and followers of followers of followers.
This would also weed out sources with articificially-inflated numbers from the Suggested Users List, because as Anil Dash says, nobody has a million followers. Typically many of these Twitter celebrities have a lot of empty storefronts following them if you dig deep enough.
Time and Keywords
Lastly, filters should be generated on the fly, dynamically, for two reasons:
1. Because experts in one category may be clueless on other subjects.
2. Because time is an incredibly valuable data point in Twitter, and expertise can change dramatically from one hour to the next depending on, for instance, the location or circumstances of a user.
We should judge a source’s expertise on a keyword by the number of retweets of her content containing that keyword over a particular time span linked to the percieved nature of a keyword — say, 3 days for "Haiti" or 30 days for "Obama".
As an example of this utility: if a user travels to Haiti from one day to the next, her authority on the topic has increased dramatically.
Conclusion
It’s just a start, but we’re missing something big in the explosion of real-time content: filters. This could be a step in evolving what we have to create a truly valuable new information resource.