XML entities (< >) appear unescaped in sahidic.ot CoNLL-U, including the lemma column

[saintsandscholars]

Three files in sahidic.ot contain XML character entities in their CoNLL-U output. CoNLL-U is a plain-text format, so < / > are carried through as literal text by any parser.

Affected files

file	occurrences
sahidic.ot/sahidic.ot_CONLLU/01_Genesis_32.conllu	3
sahidic.ot/sahidic.ot_CONLLU/01_Genesis_41.conllu	3
sahidic.ot/sahidic.ot_CONLLU/71_Sirach_05.conllu	6

12 < and 12 > in total.

Where they appear

In the # text = metadata comment, in the multiword-token form, and in both the form and the lemma columns:

19-21	ⲁϥϫⲟⲩϩ<ⲉ>	_	_	_	_	_	_	_	_
21	ϫⲟⲩϩ<ⲉ>	ϫⲟⲩϩ<ⲉ>	VERB	V	VerbForm=Fin	...

The intended reading is ⲁϥϫⲟⲩϩ<ⲉ> — the angle brackets marking an editorially supplied letter.

Because the escape reaches the lemma column, a lexical index built from this sub-corpus gets ϫⲟⲩϩ<ⲉ> as a distinct type from ϫⲟⲩϩ<ⲉ>, so occurrences of the same verb split across two lemmas.

Scope

No other sub-corpus in the repository has any entity in its .conllu files.

Reproduction

grep -rho '&[a-z]*;' --include='*.conllu' . | sort | uniq -c

[amir-zeldes]

Thanks for reporting - this will be fixed the next time this corpus is released.