Closed gobigrassland closed 5 years ago
@gobigrassland
our alphabet contains at least the letters A-Z, a space (_), and a blank token (-), where the latter is required by CTC networks. We will also be using an end-character (>) to be predicted after the last word.
In reality, it is up to you to define your alphabet depending on the actual application.
@csukuangfj Thank you very much for your reply.
in function greedy_decoder, alphabet = list(ascii_lowercase) + [' ', '>']. But in function prefix_beam_search, alphabet = list(ascii_lowercase) + [' ', '>', '%']. i feel confused.