Like I tweeted, I really can't exaggerate how important this is to hacker journalism. Nieman's introduction to data-driven approach and auto-tagging is just the first step. It's not terribly useful now--sort of a toy more than anything. There's so much work to be done, and this is, in my view, a profoundly important signal to the world of journalism to focus.
Can you give the cliff notes on this? I checked out the blog post and the actual media cloud site. I saw a swiss army knife-looking app with "Calais" emblazoned on it.
The Calais API is great, but I know what it does and what it's capable of. There are lots of apps using it. How does this extend it?
Nieman's project doesn't extend Calais--just uses it well. They say they have some homegrown analysis happening too, though they unfortunately don't specify. You'd think Nieman would consider open-sourcing its good code.
As to what the very simple service does now, though, the summary on the post does well:
— How do specific stories evolve over time? What path do they take when they travel among blogs, newspapers, cable TV, or other sources?
— What specific story topics won’t you hear about in [News Source X], at least compared to its competitors?
— When [News Source Y] writes about Sarah Palin [or Pakistan, or school vouchers], what’s the context of their discussion? What are the words and phrases they surround that topic with?
So there are some very basic bar charts and maps that are the results of these queries. The data is pretty good, though far from perfect, which @EthanZ discusses in the interview. It doesn't really help, for instance, to include the tags "united states" and "america" in the same viz as separate descriptors. Also, although I'm sure they keep the labels on the x-axis of their bar charts to themselves because it's not terrible intuitive or human-readable, it's too bad Nieman doesn't offer some description. It's really hard to compare otherwise.
Like I tweeted, I really can't exaggerate how important this is to hacker journalism. Nieman's introduction to data-driven approach and auto-tagging is just the first step. It's not terribly useful now--sort of a toy more than anything. There's so much work to be done, and this is, in my view, a profoundly important signal to the world of journalism to focus.
Can you give the cliff notes on this? I checked out the blog post and the actual media cloud site. I saw a swiss army knife-looking app with "Calais" emblazoned on it.
The Calais API is great, but I know what it does and what it's capable of. There are lots of apps using it. How does this extend it?
Nieman's project doesn't extend Calais--just uses it well. They say they have some homegrown analysis happening too, though they unfortunately don't specify. You'd think Nieman would consider open-sourcing its good code.
As to what the very simple service does now, though, the summary on the post does well:
— How do specific stories evolve over time? What path do they take when they travel among blogs, newspapers, cable TV, or other sources?
— What specific story topics won’t you hear about in [News Source X], at least compared to its competitors?
— When [News Source Y] writes about Sarah Palin [or Pakistan, or school vouchers], what’s the context of their discussion? What are the words and phrases they surround that topic with?
So there are some very basic bar charts and maps that are the results of these queries. The data is pretty good, though far from perfect, which @EthanZ discusses in the interview. It doesn't really help, for instance, to include the tags "united states" and "america" in the same viz as separate descriptors. Also, although I'm sure they keep the labels on the x-axis of their bar charts to themselves because it's not terrible intuitive or human-readable, it's too bad Nieman doesn't offer some description. It's really hard to compare otherwise.
Smart analysis of media coverage patterns? very cool.