kern
New members flow: can we do spam detection in order to let new members in faster?
We have problems adding new members.
- We get a ton of requests
- There are some people who are clearly spammers and/or bots, but most people we can't be sure of.
Possible hackbot solutions to this issue:
- Have a way for us to add new members, but notify hackbot about it.
- Hackbot can run their new posts against spam detection model. Set user to quiet and notify mods if they're flagging positive.
- Mods can deal with new user to see if actually spam. If not spam, unquiet. If spam, add data to a custom spam classifier (this isn't necessary but would be cool).
Result: we can add more members faster while maintaining good content.
I like the idea of automating on-boarding into the Group. Facebook doesn't expose membership requests via the Graph API at the moment, so we'd have to implement this through a scraper.
Could this be a job for MegBot? @maruthven
@kern This wasn't about automating onboarding. This was about making sure we catch any spammers we let in.
If someone wants to build the scraper part that could be helpful though. On Tue, Aug 25, 2015 at 8:28 PM, Alex Kern [email protected] wrote:
I like the idea of automating on-boarding into the Group. Facebook doesn't expose membership requests via the Graph API at the moment, so we'd have to implement this through a scraper.
Could this be a job for MegBot https://github.com/maruthven/megbot? @maruthven https://github.com/maruthven
— Reply to this email directly or view it on GitHub https://github.com/kern/hackbot/issues/21#issuecomment-134778101.
Jennifer Rubinovitz
Ah, misunderstood the first point under possible solutions. Yeah, this looks possible just using the Graph API. I think a general spam-detection filter is simple, straightforward, and useful (we can use the sentiment filter for reference). We can grab the member's join date, and if it's within a fixed time-frame, automatically quiet ban the user if the spam filter flags them.