Snorkel Code

[k4fk4h]

Hello, may I please view the snorkel code mentioned in this paper? Thanks in advance.

Existing bot detection datasets often rely on manual annotation or crowdsourcing, while it is labor- intensive and thus not feasible with the large-scale TwiBot-22. As a result, we adopt weak supervision learning strategy to generate high-quality labels. We firstly invite bot detection experts to annotate 1,000 Twitter users in TwiBot-22. We then generate noisy labels with the help of bot detection models. Finally, we generate high-quality annotations for TwiBot-22 with Snorkel [Ratner et al., 2017].

[BunsenFeng]

@TamSiuhin It would be great if you could find some time in your busy ICLR-chasing schedule to address this. Thank you!

[iamsreiche]

Hi, thanks a lot for your great paper and making the code available here.

The part about Snorkel caught my attention and I'm super interested in knowing how you implemented this.
Do you plan to follow-up on your previous comment or do you have some concerns regarding open-sourcing these?