def select_spans(markup):
for object in markup.objects:
type = object.type
if type == 'Person':
yield Span(object.start, object.stop, object.type)
elif type == 'Org':
spans = [_ for _ in object.spans if _.type == 'org_name']
for span in filter_overlapping_spans(spans):
yield Span(span.start, span.stop, type)
elif type in ('LocOrg', 'Location'):
spans = [_ for _ in object.spans if _.type == 'loc_name']
for span in filter_overlapping_spans(spans):
yield Span(span.start, span.stop, type)
def adapt_factru(markup):
spans = list(select_spans(markup))
# мид Грузии
# ORG-------
# LOC---
spans = list(filter_overlapping_spans(spans))
spans = list(adapt_spans(spans, markup.text, FACTRU_TYPES))
return Markup(markup.text, spans)
def adapt_ne5(markup):
# ne5 bug
# Бражский район Подмосковья
# --------------
# -----------------
# компания "Союзкалий"
# -----------
spans = list(filter_overlapping_spans(markup.spans))
spans = strip_spans_bounds(spans, markup.text, QUOTES)
spans = adapt_spans(spans, markup.text, NE5_TYPES)
return Markup(markup.text, list(spans))
def adapt_natasha(markup):
spans = list(filter_overlapping_spans(markup.spans))
spans = list(adapt_spans(spans, markup.text, NATASHA_TYPES))
return Markup(markup.text, spans)
def parse_bsnlp(record):
spans = find_spans(record.text, record.substrings)
spans = list(filter_overlapping_spans(spans))
return BsnlpMarkup(record.text, spans)
