[ddj] tools to crowdsource the news

Nicolas Kayser-Bril n.kayserbril at gmail.com
Thu Feb 9 12:30:19 UTC 2012


Hi Mehdi,

With friends in Paris we'd thought about such a tool, and it appeared that
it's extremely difficult to produce actionable information using automated
techniques or mass participation.

For similar image search, TinEye has a commercial API. 2 years ago, I
bootstraped an API for http://errorlevelanalysis.com/ combined with an EXIF
extractor and, while the tool I built proved very popular, it never helped
in uncovering a photoshopped image.

What I'd do today if I were to build a photo/video checker would be a wide
network of experts that can be activated instantly, for most of the
important information cannot be extracted automatically (sometimes, it's
the make of the car in the background of the picture that'll indicate that
it's forged, sometimes it's the error level analysis - you never know).
Similarly for an information in a broader sense: profiling the source
automatically only gets you so far. KCNA, the North Korean news agency,
would probably rank below the last level of credibility you can imagine.
Nevertheless, what it says or omits is of paramount importance to
understand the situation in DPRK. In the end, the real value-added lies in
human analysis. The platform should make it easier, faster and more
reliable.

(Of course, that doesn't hold true if you have a team of 10 MIT PhDs :-)

Best wishes and good luck

Nicolas Kayser-Bril
--
CEO and co-founder at Journalism++ <http://jplusplus.org> • @nicolaskb •
 Personal website nkb.fr <http://nkb.fr/?m> • Paris: +336 50 57 53 80 •
Berlin: +49 170 287 5332



On Thu, Feb 9, 2012 at 1:14 PM, Mehdi El Fadil
<mehdi.elfadil at mango-is.com>wrote:

> Hello,
>
> I am working on a software project to help journalists crowdsource the
> news.
>
> Does anybody know about something like that already existing? Think
> Storify or Tweetdeck but more powerful.
>
> Some features I have in mind:
>
>    - searching for similar images and determine whether an image is
>    photoshoped.
>    - profiling source of an information to guess reliability
>    - auto-detect and filter mainstream accounts in order to notice
>    relevant information produced by people less active on twitter, but maybe
>    closer to the events.
>    - ...
>
> Do newsroom already have tools dedicated to do these tasks?
>
> thanks,
>
> --
> *
>  Mehdi El Fadil
> twitter: @me_bx <http://twitter.com/me_bx> / @mango_info<http://www.twitter.com/mango_info>
> website: http://www.mango-is.com
> linkedin: http://be.linkedin.com/in/elfadme
> *
> <http://www.mango-is.com>
>
>
> _______________________________________________
> data-driven-journalism mailing list
> data-driven-journalism at lists.okfn.org
> http://lists.okfn.org/mailman/listinfo/data-driven-journalism
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/data-driven-journalism/attachments/20120209/2c36b89d/attachment-0001.html>


More information about the data-driven-journalism mailing list