tree 5557c50ac746a0fafb6c521576f48f367f42ee89
parent 517ff4b6e333c2ef5671339760ffc9a1ea2aa0f2
author Marc Kupietz <kupietz@ids-mannheim.de> 1780822610 +0200
committer Marc Kupietz <kupietz@ids-mannheim.de> 1780822610 +0200

Add stand-off metadata support for Krill output

Collect text-level metadata from stand-off XML files (one file per corpus,
in the <standOff> form produced e.g. by the wiki-taxonomy classifier) and
emit it as Krill metadata fields, joined to texts by raw_text/@docid.

- StandoffMetadata: auto-detects inputs by content (root element <standOff>),
  so no CLI option is needed; filenames are cosmetic. Classification layers
  become type:keywords fields named after the layer xml:id; links layers
  become type:attachement (KorAP-link encoding).
- By default ingest everything in the file (the annotation tool already
  applied its own top-k/threshold). The top-n/min-cert knobs are kept as
  constants/params for a future --standoff-select option.
- KrillJsonGenerator: new StandoffField type and KrillTextData.standoffFields,
  rendered after header fields; never clobbers an existing header field key.
- KorapXmlTool: detects and strips stand-off XML inputs from the ZIP list,
  parses them (krill only; warns otherwise), and injects per text.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Change-Id: Ibd50d9393c052ab8172d1e41759b19ce47380e2c