Today we (the Media Standards Trust) are launching churnalism.com, a free independent website that allows people to compare press releases with published news articles – to help identify ‘churnalism’.
It’s an idea we’ve been talking about for a good few years, but only finally got around to rolling our sleeves up last year. Ever since Nick Davies published Flat Earth News – detailing the vast quantities of press releases that make it into mainstream media – we’ve been wondering how to help distinguish churnalism from journalism.
Last spring we bit the bullet and asked Donovan Hide – a Liverpool based techno guru – to help us work out how to create it, building on the foundations of journalisted.com (which we also run). We’ve had to finance it from core funding (which we get from charitable foundations) since it’s pretty tricky to convince people to support this without seeing how it works. But now it’s up and running people will – we hope – see how useful it is and flock to our aid … www.justgiving.com/mediastandardstrust
How does churnalism.com work out if articles are churn?
When you paste a press release into churnalism.com and hit ‘compare’ the churn engine compares it with over three million articles published in the national press in the last three years (refreshed every hour or so).
The engine looks for 15-character strings in the press release that are exactly the same as 15-character strings in articles. When it finds the same string the engine looks for more identical strings in the same article. If more than 20% of the article and the press release overlap, the engine suggests it may be churn.
This makes the process sound quite simple. It’s not. For those who would like a much more sophisticated explanation of how it works, we’ll be publishing a post by Donovan Hide tomorrow.
Finding press releases
Finding press releases to compare with news articles is not as easy as it sounds. Though there are some press release aggregators that publish lots of releases on the web (like PR Newswire), this only covers a small percentage of the total number sent out. Many press releases are only published on the website of the organisation they are sent from. Many more are not published on the web at all but emailed directly to specific journalists.
This is why we’ve taken a mixed approach – part automated collection, part crowd sourcing. For the automated part, we scrape press releases from a bunch of organisations that send out lots – the government, big retailers like Tesco and M&S, and some police forces. We compare these automatically with all the articles published on national newspaper websites, and on the BBC and Sky.
For the crowdsourcing part, we allow people to paste in press releases and compare them. If they look like churn then you can save the press release (with the web link if it has one) so you can share it (e.g. via Facebook and Twitter) and so that other people can see it when they come to the site.
We’ll also be tweeting good churn from @churnalert, and building up a bank of good examples.
We have (quite a big) wishlist of other stuff we’d like to do with churnalism.com but don’t yet have the time or the money:
- More press releases collected automatically We’ll be scraping more press releases from the web as we find them. If you know of any good places to find them please let us know
- More news articles to compare With more resources we’d be able to provide comparisons of the local press as well as the national, by extending the reach of journalisted.com. Then we could cover specialist and trade press. Then international…. (OK, we’ll need to draw the line somewhere)
- Exposing probable churn by cross correlating news articles If we cross correlated news articles using the same methodology we would identify clusters of articles that overlap with one another. This could indicate that these articles may be based on press release. We could then appeal for help to find the release
- Linking churnalism.com with journalisted.com Now churnalism.com is up and running we can work out how best to link it to journalisted.com. We could, for example, indicate when a journalist’s article looks like it might be churn, and link directly from the article on journalisted.com to the press release on which it might be based
- The Daily Churn Once there are enough people using churnalism, and exposing good churn reasonably quickly, then we’ll be able publish a ‘Daily Churn’, highlighting newspapers and articles that day that appear to be churn.
Tell us what you’d add to this wishlist by emailing firstname.lastname@example.org.
All feedback much appreciated, again at email@example.com.
It hasn’t escaped our notice that there isn’t much like churnalism.com out there just now. On top of which, some of the tech we’ve developed over the past year is pretty clever. This is why, to keep churnalism.com free and independent, we’re exploring some commercial applications of the software we’ve developed.