Hacker News new | ask | show | jobs
Tealeaf (tealeaf.live)
8 points by whall 1949 days ago
2 comments

So Roberta, the deep learning model used, uses text to infer the price of stock. So if it's modeled from language what is its data source? Twitter? Reddit? How frequently does it update?
From what I understand, they're using Reddit right now, but the model will apply to any text source. I'm guessing twitter is a bit too big for now... It updates every thirty minutes.
I'm curious if the builder thought about feedback. What would happen if the users of wherever this is pulling data from start making decisions based on it's outputs?
if they have the language model they could retrain it to include the meta discussion