Existence Analysis Model, Iteration #3
After the initial test, increased the data and made an initial test to see how much score prediction changed for Iteration #2.
Then I switched from bert-case-uncased to bert-large-cased-whole-word-masking as the base model to train it from.
Switching the model out is Iteration #3, which it's more nuanced and sensitive.
Some of the test prediction scores changed little and other changed significantly.
