UDD-1 / scripts /udtools /tests /test-cases /invalid-level1 /unicode-normalization.conllu
rain1024's picture
Restructure udtools and add utility scripts
d1693da
# Unicode has multiple ways how to encode certain glyphs. UD data must always use the canonical way.
# For example, 'č' is one character (U+010D = LATIN SMALL LETTER C WITH CARON) and it is OK because it is the canonical encoding of this character.
# The non-canonical way is 'č' (two characters: U+0063 = LATIN SMALL LETTER C, U+030C = COMBINING CARON).
# On the other hand, 'с̌' (two characters: U+0441 = CYRILLIC SMALL LETTER ES, U+030C = COMBINING CARON) looks the same but it cannot be normalized to a single character because it is supposed to be Cyrillic.
# In the example below, we use the Cyrillic+Combining sequence in FORM and Latin+Combining in LEMMA. Only the latter should be reported as an error.
# sent_id = 1
# text = proс̌
1 proс̌ proč ADV _ _ 0 root _ _