nzilbb.formatter.csv
nzilbb.formatter.csv (1.4.1)
Serializer for converting to CSV files.
This formatter converts transcripts to Comma Separated Values (CSV) files.
Each row in the csv file represents a time-slot during the transcript, and each selected layer has a column for its label, start time and end time.
For each time-slot, all annotations in the selected layers that start at that time have their label appear in the corresponding column. If multiple annotations start at the same time, their labels are concatenated in the same column delimited by a newline (i.e. a multi-line cell).
e.g.
| transcript | word | word start | word end | pos | pos start | pos end |
|---|---|---|---|---|---|---|
| mop03.trs | Time | 1.0 | 1.5 | ADJ | 1.0 | 1.5 |
| mop03.trs | flies | 1.5 | 1.75 | N | 1.5 | 1.75 |
| mop03.trs | don't | 2.0 | 2.5 | AUX | 2.0 | 2.25 |
| mop03.trs | NOT | 2.25 | 2.5 | |||
| mop03.trs | like | 3.0 | 3.25 | V | 3.0 | 3.25 |
Configuration
The following parameters can be specified for the formatter:
- minimumAnchorConfidence (Min. Anchor Confidence)
- Minimum confidence for an annotation's offset before it's included in the CSV file. Generally, if set to 100, only manually set offsets will be output, if set to 50, manually set and automatically aligned offsets with be output. Specify 0 for all offsets regardless of reliability. If nothing is specified for this setting, then no start/end time columns are included in the CSV file.
- labelDuration (Label Entire Duration)
- Whether to include the label in the column for the whole duration of the annotation. If false (the default) the label will be included only on the row where the annotation starts. If true, the label will be included on every row between its start time and end time.