This requires some extra work on top of our TSV import (#1). As a common format in the field, support for CONLL-U/X would make it easy to get linguistically annotated data into STAM.