ocr-d / gt-labelling Goto Github PK

@bertsky:
I am missing some metadata for the following cases:

• the levels of ground truth
• antiqua font with or without "ſ", with "ß" or "ſs" or "ss" (e.g. https://www.deutschestextarchiv.de/book/view/goethe_metamorphose_1790?p=10)
• legal/technical/commercial/linguistic/... Special characters
• letter spaced and/or bold print
• umlaut äöü and/or aͤoͤuͤ
• old greek (greek diacritics...)

Can you add something? @bertsky

To me, it's not obvious to what certain labels pertain in particular:

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 118 to 130 in 0242e4f

  	<xsd:enumeration value="content-type/metadata/structure">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Structure of an object of some sort
  	Examples:
  	Document structure,
  	Table structure</xsd:documentation>
  	</xsd:annotation>
  	</xsd:enumeration>
  	<xsd:enumeration value="content-type/metadata/structure/toc">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Table of contents of a book, newspaper etc.</xsd:documentation>
  	</xsd:annotation>
  	</xsd:enumeration>

  → what exactly is this for?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 433 to 444 in 0242e4f

  	<xsd:enumeration value="activityDomain/computing/visual/analysisRecognition/tables">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">The recognition of table/form structure and/or contents.
  	Examples:
  	Stock exchange data in a newspaper,
  	Filled in questionaire form
  	Related:
  	OCR
  	Object / shape recognition (e.g. table separator detection)</xsd:documentation>
  	</xsd:annotation>

  → is this detection (whether, identity) or recognition (what, structure)?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Line 1264 in 0242e4f

  <xsd:enumeration value="contentOfInterest/visual/composite/tables"/>

  → "material contains tables" ?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 1322 to 1324 in 0242e4f

  	<xsd:enumeration value="granularity/logical/table/column">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Table column</xsd:documentation>

  → "annotation contains TableRegion/Roles/TableCellRole/@columnIndex" ?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 1327 to 1329 in 0242e4f

  	<xsd:enumeration value="granularity/logical/table/row">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Table row</xsd:documentation>

  → "annotation contains TableRegion/Roles/TableCellRole/@rowIndex" ?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 1332 to 1334 in 0242e4f

  	<xsd:enumeration value="granularity/logical/table/cell">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Table cell</xsd:documentation>

  → "annotation contains one TableRegion/TextRegion for each table cell" ?

• gt-labelling/xsd_schema/OCR-D_GT_schema.xsd

  Lines 278 to 285 in 0242e4f

  	<xsd:enumeration value="content-encoding/structured/tabular">
  	<xsd:annotation>
  	<xsd:documentation xml:lang="en">Content encoded in tabular form
  	Examples:
  	A tab-separated table with headings and values</xsd:documentation>
  	</xsd:annotation>
  	</xsd:enumeration>

  → does that mean: continuous text that happens to be typeset with tabs or ellipses, but does not formally constitute a table (and should not get annotated as such)?