TLDR-444 added words bbox pdfminer (#313)

* TLDR-444 added word supporting into pdfminer-reader * TLDR-444 added word extraction from pdfminer; pdfminer refactoring * TLDR-444 added tests (word bounding box) * TLDR-444 fixed code style * TLDR-444 fixed after review
ispras · Aug 29, 2023 · 436c7f8 · 436c7f8
1 parent d4c3d34
commit 436c7f8
Show file tree

Hide file tree

Showing 6 changed files with 233 additions and 162 deletions.
diff --git a/dedoc/readers/pdf_reader/pdf_txtlayer_reader/pdf_txtlayer_reader.py b/dedoc/readers/pdf_reader/pdf_txtlayer_reader/pdf_txtlayer_reader.py
@@ -8,7 +8,7 @@
 from dedoc.readers.pdf_reader.data_classes.pdf_image_attachment import PdfImageAttachment
 from dedoc.readers.pdf_reader.data_classes.tables.scantable import ScanTable
 from dedoc.readers.pdf_reader.pdf_base_reader import ParametersForParseDoc, PdfBaseReader
-from dedoc.readers.pdf_reader.pdf_txtlayer_reader.extractor_pdf_textlayer import ExtractorPdfTextLayer
+from dedoc.readers.pdf_reader.pdf_txtlayer_reader.pdfminer_reader.pdfminer_extractor import PdfminerExtractor
 from dedoc.train_dataset.train_dataset_utils import save_page_with_bbox
 
 
@@ -25,7 +25,7 @@ def __init__(self, *, config: dict) -> None:
         :param config: configuration of the reader, e.g. logger for logging
         """
         super().__init__(config=config)
-        self.extractor_layer = ExtractorPdfTextLayer(config=config)
+        self.extractor_layer = PdfminerExtractor(config=config)
 
     def can_read(self, path: str, mime: str, extension: str, document_type: Optional[str] = None, parameters: Optional[dict] = None) -> bool:
         """

diff --git a/dedoc/readers/pdf_reader/pdf_txtlayer_reader/pdfminer_reader/__init__.py b/dedoc/readers/pdf_reader/pdf_txtlayer_reader/pdfminer_reader/__init__.py