📓 Structure Ground Truth

The OCR-D structure Ground Truth contains publications in which only the structures or regions were labelled. The individual regions are marked according to the PAGE scheme.

The structure-ground truth corpus offered by OCR-D is composed of publications from the period 1500 - 1900. On the digital copy, individual regions are marked according to the PAGE scheme. In addition, individual pages are categorised according to their content.

The content of the corpus is based on manually recorded zoning data, which were collected in the course of the DFG project Deutsches Textarchiv. This data originally served to support manual transcription in the double keying process. No processing (cropping, dewarping) of the digitised data was undertaken. In contrast to the element repertoire of the PAGE format, parts of the data were indexed in greater depth within the framework of the DFG project Deutsches Textarchiv. This depth of indexing is recorded as a value of the custom attribute.

The structure ground truth can also be created in three different levels. The levels differ in scope, differentiation of the individual page types and regions. Further notes and information are presented in the following chapters:

  1. First Step : The Page Types
  2. Second Step: Page Regions
    1. Level 1
    2. Level 2