Metadata harmonisation: align spoken-data fields with UniDive naming conventions
Back to Czech PDTC · Back to index
Repo: https://github.com/UniversalDependencies/UD_Czech-PDTC
Cross-posting from the UniDive WG1 T1.5 (spoken language guidelines) metadata harmonisation review. We compared UD_Czech-PDTC’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.
1. Is the spoken portion identifiable?
This treebank mixes spoken and written material but its .conllu files don’t explicitly mark which sentences are spoken. We looked for a pattern in the data (fairly confident):
Finding: Identifiable via the newdoc id prefix pdtsc, which names a known spoken sub-corpus.
Evidence: newdoc id prefixes and counts: ln (2906), wsj (2312), pdtsc (1553), mf (1131), lnd (712), cmpr (372), vesm (209), faust (60). pdtsc is the standard abbreviation for the Prague Dependency Treebank of Spoken Czech (PDT-SC), a known spoken sub-corpus of PDTC; the rest are written-text sources (newspapers, magazines, the Wall Street Journal translation, the Faust MT-testing corpus).
Suggestion: Add # modality = spoken to all documents whose newdoc id starts with pdtsc, and # modality = written to the rest. Please confirm this reading with the PDTC maintainers, since we inferred it from the corpus name rather than internal documentation.
2. Speaker-level (naming conventions)
| Field | Suggestion |
|---|---|
global.Entity |
corpus-specific (coreference/entity annotation, project-wide) - keep, not spoken-specific |
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.