I have heard rumors of a particular count of positive examples that allowed the model to train a given word (or context of it - when talking about MLM) to be ~40. I am wondering though about the experiment setup that resulted with this result and a publication where it appeared and I have a trouble finding it. Please point me in the right direction. Is there a guideline for a minimal number of examples in the train set so that the model is able to discriminate between signal and noise?
$\begingroup$
$\endgroup$
Add a comment
|