Closed zelabean closed 2 years ago
the package currently assumes using word level lms, the code ought to work for character-based lms but chances are performance will suffer compared to word level lms. There's a bit more discussion here https://github.com/kensho-technologies/pyctcdecode/issues/50
Thank you for help : )
Is this library work with character level LM?
when i using character level lm for korean it didn't work well...