v2.2.3

• Show attached images and added ability to download attached files in the HTML output representation (API usage, return_format="html").
• Added hierarchy level information and annotations to PptxReader.

v2.2.2

• Added images extraction to ArticleReader.
• Added attachments and references to them in the HTML output representation (return_format="html").
• Fixed functionality of parameter need_content_analysis.
• Fixed CSVReader (exclude BOM character from the output).
• Added handling files with wrong extension or without extension to DedocManager (detect file type by its content).
• Update README.md.

v2.2.1

• Added fintoc structure type for parsing financial prospects according to the FinTOC 2022 Shared task (FintocStructureExtractor).
• Fixed small bugs in ArticleReader: colspan for tables, keywords, sections numbering, etc.
• Added references to nodes and fixed small bugs in the HTML output representation (return_format="html").
• Removed other_fields from LineMetadata and DocumentMetadata.
• Update README.md.

v2.2

• PdfTabbyReader improved: bugs fixes, speed increase of partial PDF extraction (with parameter pages).
• Added benchmarks for evaluation of PDF readers performance.
• Added ReferenceAnnotation class.
• Fixed bug in can_read method for all readers.
• Added article structure type for parsing scientific articles using GROBID (ArticleReader, ArticleStructureExtractor).

v2.1.1

• Update README.md.
• Update table and time benchmarks.
• Re-label line-classifier datasets (law, diploma, paragraphs datasets).
• Update tasker creators (for the labeling system).
• Fix HTML table parsing.

v2.1

• Custom loggers deleted (the common logger is used for all dedoc classes).
• Do not change the document image if it has a correct orientation (orientation correction function changed).
• Use only PdfTabbyReader during detection of a textual layer in PDF files.
• Code related to the labeling mode refactored and removed from the library package (it is located in the separate directory).
• Added BoldAnnotation for words in PdfImageReader.
• More benchmarks are added: images of tables parsing, postprocessing of Tesseract OCR.
• Some fixes are made in a web-form of Dedoc.
• Tutorial how to add a new structure type to Dedoc added.
• Parsing of EML and HTML files fixed.

v2.0

• Fix table extraction from PDF using empty config (see issue)
• Add more benchmarks for Tesseract
• Fix extension extraction for file names with several dots
• Change names of some methods and their parameters for all main classes (attachments extractors, converters, readers, metadata extractors, structure extractors, structure constructors).
  Please look to the Package reference of documentation for more details
• Add AttachAnnotation and TableAnnotation to PPTX (see discussion)
• Fix bugs in DOCX handling (see issues 378, 379

v1.1.1

• Use older pydantic version for improving compatibility with other libraries.
• Add support for RTF format.
• Fix bug in handling files' names with dots and spaces.
• Fix bug in non-integer values of text formatting in DocxReader.
• Add support of on_gpu parameter in config.
• Add attached images extraction for PdfTabbyReader.
• Fix partial file reading for PdfTabbyReader.
• Add tutorial how to create dedoc's basic data structures.
• Fix attachments_dir parameter for readers and attachments extractors.

v1.1.0

• Add BBoxAnnotation to table cells for PdfTabbyReader.
• Fix swagger, add api schema classes, remove to_dict method from ParsedDocument.
• Improve parsing PDF by PdfTxtlayerReader, add benchmarks.
• Fix BBoxAnnotation extraction for tables in PdfImageReader using table_type=split_last_column parameter.
• Change base method of metadata extractors, rename it to extract_metadata.
• Unify BBoxAnnotation extraction for all PDF readers - return only words bboxes.
• Increase timeout value for all converters.

v1.0

• Remove is_one_column_document_list parameter.
• Add tutorial about support for a new document type to the documentation.
• Improve textual layer correctness classifier.
• Improve orientation and columns classifier.
• Change table's output structure - added CellWithMeta instead of a textual string.
• Add BBoxAnnotation to table cells for PdfTxtlayerReader and PdfImageReader.
• Add ConfidenceAnnotation to table cells for PdfImageReader.
• Remove insert_table parameter.
• Added information about table and page rotation to the table and document metadata respectively.
• Use dedoc-utils library for document images preprocessing.
• Change web interface, fix online-examples of document processing.
• Add comparison operator to LineWithMeta.