v3.0.0

What's changed in 3.0.0

Major release: the code_variation column of the tsv/ files changes format, so files read with the previous schema need updating.

• code_variation column (breaking): ContainsVariation=Yes is written only for units that contain a code change; ordinary tokens have _ (the old ContainsVariation=No is no longer written). Code=, Language= and Nonce=Yes are unchanged.
• Character offsets after elisions: the position of the token features guesses, pace, volume and overlaps was shifted by one character after each elision (l'esame); values corrected in the tsv/ files.
• Empty tokens: tokens with an empty form (isolated brackets, markers attached to a space or to =) no longer exist; the transcription was corrected where needed. Token ids of the affected units were kept stable where possible.
• Stray #: tokens consisting of or containing a stray # were corrected in the transcription.
• Units made only of a marker (#_), with no content, were removed from the transcription for now; they are listed in the open issues for a decision.

.eaf, linear files, validation page and csv-metadata.json regenerated.

Full Changelog: v2.1.0...v3.0.0

v2.1.0

What's changed in 2.1.0

• EAF: new layered structure. Each speaker has the tiers ref@<speaker> (time-aligned transcription units, value = tu_id) and tx@<speaker> (enriched transcription, the only text read back). The derived word-level tiers are no longer in the files (they are generated on request). The files are valid against the official EAFv3.0 schema.
• Stable ids: unit and token ids no longer change when an .eaf is edited and synchronized back to the tsv/ file. Annotations that the .eaf cannot carry (lemma, POS, Language= labels) are kept.
• Unidentified speakers: placeholders made only of question marks (??, ???, ...) in order of first appearance in each conversation, each with a row in metadata/participants.tsv; empty form values are written as _.
• Corrected unit texts: ((P)) is now (.); stray symbols (°, >, <) attached to the neighbouring word, glued words and punctuation inside words fixed, detached eh/ah attached, breath-like isolated h fixed.
• Automatic validation of the tables against the CSVW schema on every push.

Full Changelog: v2.0.0...v2.1.0

v2.0.0

What's changed in 2.0.0

• Breaking: the topic, participants-relationship and moderator columns were removed from metadata/conversations.tsv
• Breaking: the variation column of the tsv/ files is now called code_variation and is a feature list holding all code_variation features of a token: ContainsVariation=Yes|No (unit level), Code=Other|Unsure|Underspecified, Language=<ISO code> and Nonce=Yes. It replaces the unit-level label (none/some/unspecified/all); Language and the Variation=/Orthography= features moved out of jefferson_feats
• Breaking: the unit-initial # / #_ marker is now part of the span of the unit's first token, so the original transcription can be rebuilt from the span column
• Breaking: the unit column was removed from the tsv/ files (they now have 20 columns)
• Normalization: spelling variants of interjections and discourse markers are standardized (e.g. m → mh, mhm and mmh → mhmh, he → eh, va beh → vabbè) and elision typos are corrected (e.g. dell~ → dell')
• All files are UTF-8 (no BOM), LF line endings and Unicode NFC; eaf/ is regenerated from tsv/ (audio linked by relative path)
• The 20 columns of the tsv/ files are now fully documented, and csv-metadata.json and the enriched notation reference are updated

Full Changelog: v1.1.1...v2.0.0

v1.1.1

normalize Intonation feature values

ParlaTO-transcripts-v1.1.0

• Major fix: wrong speaker attribution in linear-jefferson and linear-orthographic
• Minor fix: empty turns in linear-orthographic were removed

ParlaTO-transcripts-v1.0.0

This is the first public release of full ParlaTO transcripts. The release is also available on CLARIN at this handle
Data is released under a 4.0 CC-BY-NC-SA license.