def sort_key(cls, span: Span):
return (
-span.num_tokens,
span.match_type(),
span.offset,
span.label,
)
def sort_key(cls, span: Span):
return (
-span.num_tokens, # longest wins
span.match_type(), # exact name > exact synonym > lower case
span.offset, # deterministic
span.label, # deterministic
)
def is_keep(self, span: Span):
return (
# keep text with length > 3
len(span.text) > 3
# keep text with numbers or symbols (often measurements)
or
len(set.intersection(set(span.text), set("1234567890%<>=-+~"))) > 0
# keep text which is not solely a lowercase synonym
or span.match_type()
not in {SpanMatch.LowercaseSynonym, SpanMatch.LowercaseName})