def display_words(
self,
sentences: List[Sentence],
target: Optional[str] = None,
pdf_file: Optional[str] = None,
) -> DisplayHandle:
"""Display the bounding boxes of words.
Display the bounding boxes corresponding to words on the pdf.
"""
if not pdf_file:
pdf_file = os.path.join(self.pdf_path,
sentences[0].document.name + ".pdf")
boxes = []
for sentence in sentences:
for i, word in enumerate(sentence.words):
if target is None or word == target:
boxes.append(
Bbox(
sentence.page[i],
sentence.top[i],
sentence.bottom[i],
sentence.left[i],
sentence.right[i],
))
imgs = self.display_boxes(pdf_file, boxes)
return display(*imgs)
def get_bbox(self) -> Bbox:
"""Get the bounding box."""
if self.sentence.is_visual():
return Bbox(
self.get_attrib_tokens("page")[0],
min(self.get_attrib_tokens("top")),
max(self.get_attrib_tokens("bottom")),
min(self.get_attrib_tokens("left")),
max(self.get_attrib_tokens("right")),
)
else:
return None
def get_box(span: SpanMention) -> Bbox:
"""Get the bounding box."""
warnings.warn(
"get_box(span) is deprecated. Use span.get_bbox() instead.",
DeprecationWarning,
)
return Bbox(
min(span.get_attrib_tokens("page")),
min(span.get_attrib_tokens("top")),
max(span.get_attrib_tokens("bottom")),
min(span.get_attrib_tokens("left")),
max(span.get_attrib_tokens("right")),
)
def get_bbox(self) -> Bbox:
"""Get the bounding box."""
# TODO: this may have issues where a sentence is linked to words on different
# pages
if self.is_visual():
return Bbox(
self.page[0],
min(self.top),
max(self.bottom),
min(self.left),
max(self.right),
)
else:
return None
def _coordinates_from_HTML(
self, page: Tag, page_num: int
) -> Tuple[List[Tuple[Tuple[int, int], str]], Dict[Tuple[int, int],
Bbox], ]:
pdf_word_list: List[Tuple[Tuple[int, int], str]] = []
coordinate_map: Dict[Tuple[int, int], Bbox] = {}
block_coordinates = {}
blocks = page.find_all("block")
i = 0 # counter for word_id in page_num
for block in blocks:
x_min_block = int(float(block.get("xmin")))
y_min_block = int(float(block.get("ymin")))
lines = block.find_all("line")
for line in lines:
y_min_line = int(float(line.get("ymin")))
y_max_line = int(float(line.get("ymax")))
words = line.find_all("word")
for word in words:
xmin = int(float(word.get("xmin")))
xmax = int(float(word.get("xmax")))
for content in self.separators.split(word.getText()):
if len(content) > 0: # Ignore empty characters
word_id = (page_num, i)
pdf_word_list.append((word_id, content))
coordinate_map[word_id] = Bbox(
page_num,
y_min_line,
y_max_line,
xmin,
xmax,
)
block_coordinates[word_id] = (y_min_block,
x_min_block)
i += 1
# sort pdf_word_list by page, block top then block left, top, then left
pdf_word_list = sorted(
pdf_word_list,
key=lambda word_id__: block_coordinates[word_id__[0]] +
coordinate_map[word_id__[0]][1:3],
)
return pdf_word_list, coordinate_map