torch-rnn icon indicating copy to clipboard operation
torch-rnn copied to clipboard

Dealing with unseen characters?

Open vinhqdang opened this issue 9 years ago • 0 comments

Hello

What is the best way to deal with unseen characters which appear in start_text but not in training data?

A way I could imagine is just remove them from start_text, or replace them with a random character from the training data, but I am not sure it is a good approach.

vinhqdang avatar May 20 '16 07:05 vinhqdang