PAGEXML

PAGEXML is an XML-based data format designed to encode the layout, content, and metadata of digital page images for optical character recognition (OCR) and document analysis. It provides a portable, interoperable representation for ground-truth annotations and evaluation across digitization projects and research.

The schema represents page structure in a hierarchical manner, allowing pages to be subdivided into regions and blocks that contain text lines, words, and potentially non-text elements. Each element carries geometric information describing its position on the page, and the structure supports reading order, baselines, and other typographic features needed for layout analysis. In addition to layout data, PAGE XML can include language, font or style hints, and various metadata. The format is designed to be extensible, enabling project-specific attributes without breaking interoperability. Validation against a formal schema is possible to ensure consistency across datasets.

PAGE XML is widely used in OCR research and historical-document digitization. It supports ground-truth annotation workflows and evaluation pipelines, and it is often adopted as a common interchange format among tools and projects. The format is maintained by a community of researchers and institutions and is commonly employed in initiatives such as OCR-D and related evaluation efforts.