def from_standoff(cls, text, annotations, sentence_split=True,
discont_rule=None, overlap_rule=None,
filter_types=None, exclude_types=None,
tokenization_re=None, document_id=None):
"""Return Document given text and standoff annotations."""
# first create a document from the text without annotations
# with all "out" tags (i.e. "O"), then re-tag the tokens based
# on the textbounds.
textbounds = parse_textbounds(annotations, discont_rule)
document = cls.from_text(text, sentence_split, textbounds,
tokenization_re=tokenization_re)
if document_id is not None:
document.id = document_id
if filter_types:
textbounds = filter_textbounds(textbounds, filter_types)
if exclude_types:
textbounds = filter_textbounds(textbounds, exclude_types,
exclude=True)
verify_textbounds(textbounds, text)
textbounds = eliminate_overlaps(textbounds, overlap_rule)
retag_document(document, textbounds)
return document
def from_standoff_to_spert(cls,
text,
annotations,
sentence_split=True,
discont_rule=None,
overlap_rule=None,
filter_types=None,
exclude_types=None,
tokenization_re=None,
document_id=None):
"""Return Document given text and standoff annotations."""
# todo: убрать костыль с исключением отношений
textbounds, relations = parse_textbounds(
annotations, discont_rule, exclude_relations=exclude_types)
document = cls.from_text(text,
sentence_split,
textbounds,
tokenization_re=tokenization_re)
if document_id is not None:
document.id = document_id
if filter_types:
textbounds = filter_textbounds(textbounds, filter_types)
if exclude_types:
textbounds = filter_textbounds(textbounds,
exclude_types,
exclude=True)
verify_textbounds(textbounds, text)
relation_dict = {
arg1: (relation, arg1, arg2)
for (relation, arg1, arg2) in relations
}
docs = convert_documents(document, textbounds, relation_dict)
return docs