Take the spaces from the bounding boxes

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "pdf-document-layout-analysis"
-version = "2024.12.17.3"
+version = "2024.12.17.4"
 description = "This tool is for PDF document layout analysis"
 license = { file = "LICENSE" }
 authors = [{ name = "HURIDOCS" }]

src/pdf_features/PdfToken.py

@@ -44,11 +44,11 @@ def from_poppler_etree(page_number: int, xml_tag: ElementBase, pdf_font: PdfFont
         else:
             tag_id = "tag"
 
-        content = "".join(xml_tag.itertext()).strip()
         reading_order_no = int(xml_tag.attrib["reading_order_no"]) if "reading_order_no" in xml_tag.attrib else -1
         bounding_box = Rectangle.from_poppler_tag_etree(xml_tag)
         token_type = TokenType.TEXT
 
+        content = "".join(xml_tag.itertext()).strip()
         return PdfToken(page_number, tag_id, content, pdf_font, reading_order_no, bounding_box, token_type)
 
     def get_label_intersection_percentage(self, label: Label):

src/pdf_features/Rectangle.py

@@ -18,11 +18,20 @@ def __init__(self, left: int, top: int, right: int, bottom: int):
 
     @staticmethod
     def from_poppler_tag_etree(tag: ElementBase) -> "Rectangle":
+        content = "".join(tag.itertext())
+
         x_min = int(tag.attrib["left"])
         y_min = int(tag.attrib["top"])
         x_max = x_min + int(tag.attrib["width"])
         y_max = y_min + int(tag.attrib["height"])
 
+        one_character_length = max(int((x_max - x_min) / len(content)), 2)
+        if content[0] == " ":
+            x_min += one_character_length
+
+        if content[-1] == " ":
+            x_max -= one_character_length
+
         return Rectangle(x_min, y_min, x_max, y_max)
 
     def fix_wrong_areas(self):