propublica / facebook-political-ads

Monitoring Facebook Political Ads
MIT License
237 stars 50 forks source link

Collecting ads in a new country #79

Closed cjer closed 6 years ago

cjer commented 6 years ago

I would like to make a campaign for collecting political ads in Israel (I have been testing the Firefox and Chrome extensions myself for the past few weeks). How would you go about doing so? Is there a way to gain access to the raw data collected by the original extensions? Will it be better to fork, build and run on a local environment?

Much obliged

jeremybmerrill commented 6 years ago

Hi @cjer,

Heh, do you anticipate elections soon?

We're not adding new international partners at this time, since the US elections are approaching, so, yes, forking the project entirely (this repo, plus, the https://github.com/propublica/fbpac-api-public repo for the admin dashboard in which you'd be able to see and browse the collected data) would be the best option. You'd have to upload your own version of the extension for Mozilla and Chrome's extension stores, too.

You'd need to make some small-ish changes in https://github.com/propublica/facebook-political-ads/blob/master/extension/src/parser.js#L572-L589 to detect whatever the string is that identifies ads in Hebrew. (It's "Sponsored" in English, "Sponsorisé" in French, etc.) You can also add a translation for the extension into Hebrew here: https://github.com/propublica/facebook-political-ads/tree/master/extension/_locales

Feel free to send me an email if you have questions...

(The data you (or any other Hebrew-speaking participants) submitted has ended up in my database. While there's no pre-trained model to distinguish Hebrew political ads from non-political ads, here's an example I found of what a political ad looks like in the admin:

screenshot 2018-09-03 12 12 53

I'm assuming that your Facebook is set to be in English, hence why those ads are detected by saying "Sponsored" (or "SpSonSsoSred"). Any ads by for someone whose Facebook is set to Hebrew wouldn't be detected. )