Hi, If my understanding is correct, this flag uses the same logic/architecture that -use_chars_enc does. The use_chars_enc uses a char-based CNN to generate an embedding matrix of size |V| x h (i.e. the input for the lstm).
But for the target side to be character based, don't we need the output matrix of size h x |V| (and not an embedding matrix) to be generated by char-based CNN?? By looking at the code it looks like this option does does not create an output matrix using char-CNN. So what exactly does it do?
Hi, If my understanding is correct, this flag uses the same logic/architecture that
-use_chars_encdoes. Theuse_chars_encuses a char-based CNN to generate an embedding matrix of size |V| x h (i.e. the input for the lstm).But for the target side to be character based, don't we need the output matrix of size h x |V| (and not an embedding matrix) to be generated by char-based CNN?? By looking at the code it looks like this option does does not create an output matrix using char-CNN. So what exactly does it do?
Any clarification would be very helpful.