Bad Idea du Jour: A Filtering Service for Social Media Channels
June 21st, 2010, 2 Comments »
I have a small problem. I really like the World Cup. Yet the games are played in the morning, with the last game finishing up shortly after lunch. I sometimes watch one game in the morning, but I have to record the rest and watch later in the day as time permits. I also occasionally do this for hockey games, particularly eastern ones that begin at 4:00pm here on the West Coast.
Of course, I spend a lot of time online, working and playing in the real-time, flow-oriented social web. So there’s a high risk of my learning the outcome of sports events before I get to watch them. I’ve heard similar complaints from people who time-shift television shows–the finale of Lost, for example, or the season premiere of True Blood.
I address this problem by going very light on Twitter, Facebook and, uh, high-risk blogs until I’ve watched whatever I recorded.
Smart Filters to Avoid Disappointment
There are various apps which offer muting functionality for individual keywords or users. What I could really use is a view of Twitter and Facebook that magically removes all messages related to, say, the World Cup.
How would we achieve this? The simplest route would be using bundles of related keywords as a filter, maybe gathered through a crowdsourced process. For the World Cup, we might block all country names and team nicknames for starters. Then maybe common terms like ‘goal’, ‘keeper’ and so forth. Next you’d probably want to block all player names. This presents an immediate problem, as you’re filtering out a bunch of common names like Lee, Kim, James and Green.
I asked about this on Twitter, and Dave Johnson suggested that it might be a good task for Google’s Prediction API.
Ideally, I guess you want a service that can algorithmically discern between “Blimey, England keeper Robert Green concedes an easy goal” and “Blimey, our England office is never going to make our goal of going green this quarter.” Presumably the service would track a user’s historic data, too, and adjust the prediction based on the likelihood that they’re talking about soccer.