Fix separators for word tokenizer #935
Description
Currently, the separators property for WordTokenizer is a string, that is parsed as a comma-separated list of characters. It was appropriate for command-line parsing, but not appropriate for API use.
A better solution is to have a char[].