What is a good sample size for retraining individual words? I retrained using 50 good and 50 bad examples for a word but was hoping that a smaller sample size would also still be efficient?
1 Like
I think it depends greatly on the size of the model, but with a small model, it seems possible to teach one word with a dataset of around 200. If all goes well, it seems that less than 500 sentences may be enough to train one word.
1 Like