Community

Regarding PDF/UA output, we would like to emphasize not only on tool pass/fail, but on semantic usability in assistive technologies particularly reading order, heading hierarchy, list/table tagging, language declaration, and consistent handling of figures. For scanned/OCR inputs, limitations are recognized; however, improved paragraph tagging, heading detection/hierarchy, and optional bookmark generation would significantly reduce tender challenges. 

Here is detailed explanation of the requirements and importance. The level A requirements are listed in the summary above.

A. PDF/UA - Core “Hard” Requirements (ISO 14289-1 aligned)

These are the items most likely to trigger customer challenge if missing:
・Document is correctly tagged and recognized as PDF/UA
・Logical reading order is correct and consistent
・Headings use a valid, hierarchical structure (H1-H6)
・All meaningful content is tagged (no untagged paragraphs)
・Images with meaning have alternative text
・Decorative elements are marked as artifacts
・Language is declared at document level
・Tables are correctly structured (headers, scope, associations)

Gaps in these areas are typically treated as material non-compliance risks in tender evaluations, regardless of tool output.

B. Common “Warnings”/Judgement-based Areas

These are frequently acceptable if reasonably justified, especially for scanned/OCR source documents:
・Headings present without bookmarks
・Multiple H1 elements (context-dependent)
・OCR limitations where no semantic source exists
・Alt-text quality where content is purely informational

These are usually handled as “supported with limitations”, not blockers.

C. ABBYY/Scanned Document Context (Important for Positioning)

Given scanned input, it’s understood (in both EU and NA evaluations) that:
・semantic data is limited vs born-digital sources, and
・full automation of high-quality alt-text is not currently achievable.

The practical expectation is standards-aligned output with known limitations, not perfection.