Hey @curators
you might have noticed something new on the homepage of Phoenix, Docbot has started to create points!
Q: Who is Docbot?
A: Docbot is a machine learning algorithm developed by Evan and will help us read ToS! Forcing robots to read terms so we don’t have to. Amazing!
Q: No more humans?
A: No. Points created by Docbot will require manual review by a curator or higher like points created by humans do.
Q: Can I start accepting Docbot’s points?
A: Yes, if you have the necessary permissions (curator or higher). But please be careful when reviewing Docbots points! Like humans, robots make mistakes and that is okay.
Q: What has Docbot analyzed so far?
A: As of 3/25 it has gone through all English documents in the ToS;DR database, looking for evidence of 19 different cases. In the coming weeks we will run it on around 100 additional cases.
Q: Do I need to report incorrect Docbot points?
A: Typically no, you can just reject the submission and future versions of Docbot will use this signal to improve.
Q: What if Docbot submitted multiple points for the same case for a service?
A: This should only happen if it found evidence of the case in multiple documents for that service. In this situation, you can approve both (assuming they are both correct)
Q: When will Docbot be advertised publicly?
A: Soon! First we need your help to see if this hybrid workflow works as intended. General feedback on output quality or the curation process would be very helpful
That’s all for right now, thank you for your help and happy curating