home edit page issue tracker

This page pertains to UD version 2.

It appears that you have Javascript disabled. Please consider enabling Javascript for this page to see the visualizations.

Abaza ATB

Overview


type	only spoken
available since	2.11
link	https://github.com/UniversalDependencies/UD_Abaza-ATB
genre	spoken
contact	alexeykochevoy@gmail.com
sentences	98
tokens	652

Issue draft: UD_Abaza-ATB

Is spoken part clearly identifiable? N/A

Every sentence carries exactly the same six comment fields, plus one MISC feature on (nearly) every token:

Field	Level	Example	Standard name (metadata.md)	Advice
`text_orth`	sentence	`# text_orth = сарА сы-хьиз фатима-пI`	this seems a morpheme-segmented orthographic form (hyphens mark morpheme boundaries, stress marked)	rename to `text_morphemic`
`text_transcription`	sentence	`# text_transcription = sará sə-χ'iz fatima-ṗ`	Latin-script rendering of the Cyrillic orthographic form (not a phonetic/IPA transcription)	rename to `text_transliteration`
`text_rus`	sentence	`# text_rus = Меня зовут Фатима.`	translation field; project convention uses 3-letter ISO codes	OK
`text_name`	sentence (but constant across all sentences of one recording — 6 distinct values over 98 sentences)	`# text_name = Professija_AjsanovaFB_11072017_checked.eaf`	this seems a document identifier, wrongly repeated per-sentence instead of set once per document	convert to `# newdoc id = ...` at the first sentence of each of the 6 documents (drop the per-sentence repetition and the `.eaf` extension)

Speaker metadata: none at all (speaker_id, speaker_role, etc. absent). The 6 text_name filenames encode a speaker per recording (e.g. AjsanovaFB, SanashokovaCKh, DzhuzhuevKM, BidzhevaTA, AsanaevaFM — initials suggest one speaker/informant per file). Once text_name is converted to newdoc id, a speaker_id could plausibly be derived from the same filename component.
Genre: no # genre = ... field, even though topics are recoverable from filenames. These read as personal narrative/interview elicitations — could add # genre = narrative or interview per document.
sound_url: the corpus homepage (lingconlab.ru/spoken_abaza) implies underlying audio recordings exist; no link is included in the .conllu. Worth asking whether individual recording URLs can be shared.