nzilbb.formatter.trmparsercsv (0.1.1)

Generates CSV files specificially for export/import of data for the trm-parser implemented by Connor Talyor-Brown for Māori data.

When serializing fragments, the following transformations are made:

• Vowels with umlauts or followed by a colon are macronized.
• English words are enclosed in square brackets.
• Utterances are split on full stops and pauses of 1000ms or longer, creating two fragments per utterance.
• All punctuation is removed.

A CSV file is generated with the following columns:

• Document - the transcript ID.
• Speaker - the participant ID.
• MatchId - the MatchId-encoded identifier for the fragment.
• ID - the unique identifier for the fragment.
• Original - the original, unstandardized text of the fragment.
• WithPauses - the standardized fragment text with pause length (in seconds) between each token.
• Terminator - the reason for terminating the fragment, which can be:
  • . or - : there was a pause marker,
  • A number like 1.234 : there was an inter-token pause,
  • utterance : it was the end of the utterance, or
  • turn : it was the end of the speaker turn.
• Fragment - the standardized fragment text.