def _process_mm_biomarker_result(spans: List[Span],
tokens: List[DocToken]) -> List[Span]:
result_spans = []
measurement_spans = [x for x in spans if x.label == 'MEASUREMENT']
if len(measurement_spans) > 1:
abnormal_text = measurement_spans[0].text
total_text = measurement_spans[1].text
result_spans.append(
Span(text=abnormal_text,
entity=AbnormalCells.create(abnormal_text),
tokens=measurement_spans[0].tokens))
result_spans.append(
Span(text=total_text,
entity=TotalCells.create(total_text),
tokens=measurement_spans[1].tokens))
elif len(measurement_spans) == 1:
total_text = measurement_spans[0].text
result_spans.append(
Span(text=total_text,
entity=TotalCells.create(total_text),
tokens=measurement_spans[0].tokens))
mm_mutation_type_spans = [
x for x in spans if x.label == 'MM_BIOMARKER_MUTATION'
]
if len(mm_mutation_type_spans) > 0:
token_start = mm_mutation_type_spans[0].offset
mm_mutation_type_tokens = tokens[token_start:]
text = ' '.join([x.token for x in mm_mutation_type_tokens])
result_spans.append(
Span(text=text,
entity=MmBiomarkerMutation.create(text),
tokens=mm_mutation_type_tokens))
return result_spans
def spans(doc, tokens):
spans = [
Span(
text="a",
doc=doc,
entity=Entity(name="A", label="LABEL_0"),
tokens=tokens[:1],
),
Span(
text="a",
doc=doc,
entity=Entity(name="B", label="LABEL_0"),
tokens=tokens[:1],
),
Span(
text="a",
doc=doc,
entity=Entity(name="A", label="LABEL_1"),
tokens=tokens[:1],
),
Span(
text="a",
doc=doc,
entity=Entity(name="C", label="LABEL_0", synonyms=["a"]),
tokens=tokens[:1],
),
]
assert 4 == count_it(spans)
return spans
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 _process_mm_biomarker(spans: List[Span]) -> List[Span]:
result_spans = []
if len(spans) > 0:
text = '/'.join([x.text for x in spans])
tokens = [token for span in spans for token in span.tokens]
biomarker_span = Span(text=text,
entity=Biomarker.create(text),
tokens=tokens)
biomarker_result_span = Span(
text=text,
entity=BiomarkerResult.create(biomarker_span),
tokens=tokens)
result_spans.append(biomarker_result_span)
return result_spans
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})
def _resolve_entity(self, prefix: Token, doc_tokens: List[DocToken]):
any_found = False
while not any_found and prefix:
entities = self.resolver.resolve(term=prefix)
for entity in entities:
span = Span(
text=prefix,
doc=self.doc,
entity=entity,
tokens=doc_tokens,
)
self.spans.append(span)
any_found = True
if not any_found:
prefix = prefix.left_token
doc_tokens = doc_tokens[:-1]