def test_should_add_paragraphs_without_title(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_2))
]))
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
assert section.get_paragraph_text_list() == [
SECTION_PARAGRAPH_1, SECTION_PARAGRAPH_2
]
def test_should_raw_table_for_table_text_to_section(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<table>', LayoutBlock.for_text(SECTION_PARAGRAPH_2))
]))
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
assert section.get_paragraph_text_list() == [SECTION_PARAGRAPH_1]
assert section.get_text_by_type(
SemanticRawTable) == SECTION_PARAGRAPH_2
def test_should_add_note_for_other_text_to_body(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('O', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_2))
]))
parent_section = SemanticSection(semantic_content_list)
assert parent_section.get_notes_text_list('fulltext:other') == [
SECTION_PARAGRAPH_1
]
sections = parent_section.sections
assert len(sections) == 1
assert sections[0].get_paragraph_text_list() == [SECTION_PARAGRAPH_2]
def test_should_add_separate_section_label(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<section>', LayoutBlock.for_text('1 ' + SECTION_TITLE_1)),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1))
]))
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
semantic_headings = list(section.iter_by_type(SemanticHeading))
assert len(semantic_headings) == 1
assert semantic_headings[0].get_text_by_type(SemanticLabel) == '1'
assert semantic_headings[0].get_text_by_type(
SemanticTitle) == SECTION_TITLE_1
def test_should_include_reference_citation_in_paragraph(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<citation_marker>', LayoutBlock.for_text('Ref 1')),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_2)),
]))
LOGGER.debug('semantic_content_list: %s', semantic_content_list)
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
reference_citations = list(
section.iter_by_type_recursively(SemanticReferenceCitation))
assert len(reference_citations) == 1
assert reference_citations[0].get_text() == 'Ref 1'
def test_should_include_citation_in_paragraph(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<citation_marker>', LayoutBlock.for_text(SECTION_PARAGRAPH_2)),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_3)),
]))
LOGGER.debug('semantic_content_list: %s', semantic_content_list)
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
assert section.get_paragraph_text_list() == [
' '.join([
SECTION_PARAGRAPH_1, SECTION_PARAGRAPH_2, SECTION_PARAGRAPH_3
])
]
def test_should_include_multiple_raw_equations_without_label(self):
semantic_content_list = list(FullTextSemanticExtractor(
).iter_semantic_content_for_entity_blocks([
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_1)),
('<equation>', LayoutBlock.for_text('Equation 1')),
('<equation>', LayoutBlock.for_text('Equation 2')),
('<paragraph>', LayoutBlock.for_text(SECTION_PARAGRAPH_2)),
]))
LOGGER.debug('semantic_content_list: %s', semantic_content_list)
assert len(semantic_content_list) == 1
section = semantic_content_list[0]
assert isinstance(section, SemanticSection)
raw_equations = list(
section.iter_by_type_recursively(SemanticRawEquation))
assert len(raw_equations) == 2
assert raw_equations[0].get_text() == 'Equation 1'
assert raw_equations[1].get_text() == 'Equation 2'
def get_semantic_extractor(self) -> FullTextSemanticExtractor:
return FullTextSemanticExtractor()