tree 7e54118a21a34c2ed73d9fcf7fa5ac48e26ad25a
parent 2cb3acb1fcc911409862f4f23114b30c47c640c3
author Marc Kupietz <kupietz@ids-mannheim.de> 1781465148 +0200
committer Marc Kupietz <kupietz@ids-mannheim.de> 1781467034 +0200

Index extra inline w-attributes (norm, orig, phon, trans) as Krill layers

TEI-derived corpora can carry inline <w>-level annotations beyond pos and
lemma. Read norm, orig, phon and trans from the correspondingly named <f>
features in morpho.xml and emit each as its own Krill token layer
(foundry/<name>=tokens with foundry/<name>:<value> terms, original casing
preserved), grouped under the morpho foundry — just like pos (p) and lemma
(l). The set is intentionally restricted to these four names, and the
layer-detection scan is gated behind a single short-circuiting check, so
corpora without them pay at most one extra pass. The layers are Krill-only
and do not affect CoNLL-U output.

The four new MorphoSpan fields also had to be carried through the
field-selective copy/merge in collectKrillMorphoData{,Direct}, which
otherwise silently drops them from Krill while CoNLL-U still works.

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

Change-Id: I2695acd47f33de782045f4cee50799538e6a3c4b