parsers should return an object with {'key': [[seg1_tok1, seg1_tok2, ...]]}
in the future we may find usecases where we do not need the list of lists representation (i.e. one field for each token in each segment, but for now this is the only usage we support.
this representation captures: words, pos tags, alignments
parsers should return an object with {'key': [[seg1_tok1, seg1_tok2, ...]]}
in the future we may find usecases where we do not need the list of lists representation (i.e. one field for each token in each segment, but for now this is the only usage we support.