Language detection, page orientation detection and OCR character recognition have been improved for non Latin languages. To include these improvements please update required files using traineddata.zip from Solid Framework downloads.

Next release expected on 25 November 26

Improvements: 

• [json]   Support superscript and subscript elements.
• [docx]  Improved bullet detection.
• [docx]  Improved the automatic recovery of various incorrectly encoded symbols.
• [docx]  Improved the detection, validation and correction of incorrectly encoded ligatures.
• [docx]  Improved the detection of non-orthogonal segments in table border paths.
• [docx]  Improved detection of certain Japanese characters.   
• [docx]  Improved cell detection of hybrid tables.
• [docx]  Support the detection of Microsoft Sensitivity Information Protection (MSIP) labels.
• [office] Improved the processing and conversion time of documents with large numbers of graphics.

Bugfixes: 

• [pptx]   Fixed a bug preventing the accurate detection of one line of text on a document.
• [docx]  Fixed a bug resulting in the misdetection of two characters.
• [docx]  Fixed a bug causing a single character to be duplicated in conversion output.
• [pptx]   Fixed a bug resulting in glyphs being rasterized as an image instead of preserved as text.
• [docx]  Fixed a bug causing the line spacing of an empty footer to be less than one.
• [docx]  Fixed a text box boundary issue resulting in a page number being partially clipped.