Metadata harmonisation: align spoken-data fields with UniDive naming conventions
Back to Slovenian SST · Back to index
Repo: https://github.com/UniversalDependencies/UD_Slovenian-SST
Cross-posting from the UniDive WG1 T1.5 (spoken language guidelines) metadata harmonisation review. We compared UD_Slovenian-SST’s current CoNLL-U metadata against the proposed naming conventions (see also the full treebank status table). This is a suggestion for maintainers to review - feel free to push back on anything that doesn’t fit the corpus. The comparison was carried out semi-automatically with the help of Claude (Anthropic); errors or misunderstandings are possible, so please double-check anything unclear.
1. Document-level (naming conventions)
| Field | Suggestion |
|---|---|
sound_url |
possibly move to document level |
2. Speaker-level (naming conventions)
| Field | Suggestion |
|---|---|
speaker_id |
OK |
Implementation notes
Needs manual input from maintainers
sound_url→ document level: dry-run ofpython3 workgroups/spoken-data/scripts/harmonize_metadata.py hoist-to-doc DIR --key sound_urlagainst the realtrainfile shows it’s not safe to hoist as-is - only 34 of 277 documents have a constantsound_urlacross all their sentences; the other 243 vary sentence-to-sentence (each sentence likely points at its own audio clip within the recording, not one file per document). This confirms the draft’s “possibly” hedge - please clarify whethersound_urlis meant to be per-sentence (in which case no change is needed, it’s already correctly scoped) or whether there’s a separate constant per-document URL to add instead. No script change is safe to propose until that’s settled.
This issue was prepared as part of the UniDive WG1 T1.5 spoken language guidelines effort. Happy to help implement these changes ourselves if that’s easier than doing it on your end - just let us know.