While reading Erick Schonfeld’s post about how It’s Time to Hide The Noise, in reference to how twitter clients still fail to float interesting tweets to the top, I started thinking about some features that would help get this process started.
Since I’m too busy (ok, lazy) to build any of these myself, I figured I’d just write about some features that I would definitely use and that would be more than enough for me to switch twitter clients (again). I use brizzly right now (switched from Tweetdeck), which does a great job, but still doesn’t do much to help filter the noise.
Smarter Categories:
Most clients let you categorize the people you follow, including Twitter itself with their forthcoming “lists” feature. This is somewhat useful but the problem is that the people I follow tend to have incredibly insightful tweets amidst a sea of inane ones. Right when I’m about to unfollow someone after stumbling through their stream of drivel, BOOM they send out a link to an awesome article or blog post that I otherwise wouldn’t have found. I’ve learned so many great things from the people I follow that it justifies wasting the time it takes to get through the notes about what they’re eating, when their “wheels are down” and other empty nuggets of the sort.
This wouldn’t be easy, but I think it would be doable: why not let us categorize tweets? Build your own list of categories or select from a pre-set list. Then, as you use the client and come across a tweet worth-while tweet, categorize it as such. The engine would then analyze the following characteristics about the message:
- Who it came from
- What it linked to (follow the shortener link to see which news source or blog it referenced)
- What kind of media was in the link? (Image, video, article, flash, whatever)
- Is it being retweeted? If so, are other people I follow retweeting (that may imply it is more relevant to me)?
- Are other twitter users I follow mentioned in the tweet?
The engine could then learn that I tend to think articles on TechCrunch, SFIst, LifeHacker, Consumerist, NYT, WSJ, etc are useful and thus should be somehow conveyed to me as more important. It will also identify the users that tend to send more interesting tweets.
Lexicon Analysis:
This feature would definitely prove to have some false positives (and vice-versa) but would nonetheless be worthwhile to try out and improve with time. Sort of like a spam filter, the user could create a list of words or phrases that would be used to filter the stream. Mine would surely include:
- “flight to”
- “wheels down”
- “lunch”
- “eating”
“Don’t Like”:
Like facebook’s “like” button, a great twitter client would have a “don’t like” button that could be pressed for any given tweet in the stream. The engine would start to then find patterns in the tweets you don’t like which would then suggest or at least help you build rules that could be set up to better manage the stream.
My twitter stream gets almost as much attention as my email stream does and is just about my most important news source. I think twitter clients have a long way to go before managing twitter is as efficient as managing email, and these features would help get us there.
