As we did some of the initial research for Argo, we spoke to several brilliant individuals who’d helped build sites that touch on aspects of our model. One of those people was Megan McCarthy, human editor of the media news aggregator site MediaGazer. Today that site got a shoutout from the Nieman Journalism Lab, which noticed that MediaGazer had leaped into its list of top 5 traffic referrers, after having existed for only two months.
What’s so special about MediaGazer? You might have noticed that I referred to Megan as the site’s “human editor.” That’s because the engine that powers the site is a sophisticated algorithm. Given a list of hundreds of sources around a topic, the algorithm constantly digests stories from those sources. It identifies which stories seem to have made a splash among that selectively-defined crowd. Then it clusters those stories and the reaction to them, and orders them by importance. It works a little like Google News, but thanks to its human inputs – the human-compiled source list and Megan’s constant feedback – it’s even better for getting a sense of the hot stories developing on a topic at any instant.
MediaGazer’s part of a family of sites that includes TechMeme (tech) and Memeorandum (politics). When the other sites were launched, they were purely algorithmic. But Gabe Rivera – the creator of the algorithm that powers these sites – discovered that the sites could be slow to surface a new story. In the age of Twitter, a story can become big news before a critical mass of bloggers or news organizations have written about it. So Rivera brought Megan on board to add human news judgment and speed to the algorithm’s decisions, to train it day by day and make the sites better.
We often say that “smart aggregation” is one of the three planks of the model we’re developing with Argo (the other two planks being the blogger’s guidance and the voice of the crowd). It’s also the area with the least amount of best practices and general knowledge to draw on. Models like Mediagazer show us what’s possible.
