Columbia Games Corpus » ToBI Labeling Guidelines

0. Files and Links

ToBI Documentation

• » ToBI Home Page
• » Guidelines for ToBI Labeling (PDF). Detailed manual with examples and exercises. Ignore the first section (0. Preface).
  All examples and exercises are available in WaveSurfer, after installing the packages as explained below.
  
• » The ToBI Annotation Conventions (PDF). Quick reference guide.
  

1. Getting started

1. Go to the File Locking page.
2. Look for the wav file you want to label and check the following: its "Orthographic Transcription" and "Word Alignment" tasks must have "finished" status, and its "ToBI Labeling" task must have "not started" status and no person assigned to it.
   If these conditions are not met, do not work on this file.
3. Click "EDIT" in the "ToBI Labeling" task of the wav file you want to label.
4. On the next screen, change its status to "in progress", and enter your name.

2. Getting the annotation files

• In the "ToBI Labeling" column of the File Locking page, you will notice that under the status label there are links to the several files.

• WaveSurfer

  • Save a copy on your computer of the ".wav", ".words", ".breaks" and ".misc" files (right-click on each of them and choose Save target as..., Save link as..., or similar).
    Important: Do not change the name of these files, since they must match exactly the name of the .wav file. Example: "s09.objects.1.A.wav", "s09.objects.1.A.words", "s09.objects.1.A.breaks", "s09.objects.1.A.misc".
  • Now you are ready to begin labeling with WaveSurfer. Put all four files (".wav", ".words", ".breaks", ".misc") in the same directory, and open the ".wav" file from WaveSurfer, choosing the "ToBI-Games" configuration.
  • The ".breaks" file corresponds to ToBI's breaks tier, and has one X label at each word boundary. To change a label, click on the X and replace it with a new label using the keyboard.
  • The ".misc" file corresponds to the miscellaneous tier, with labels for self-repairs, coughs, laughs and breaths. This tier should be already labeled, but if you find things not correctly labeled, feel free to update it.
  • The ".words" file corresponds to the orthographic tier. There should be not need to modify this file, but feel free to update it if you do not agree with the orthographic transcription or alignment.

• Praat

  • Save a copy on your computer of the ".wav" and ".TextGrid", files (right-click on each of them and choose Save target as..., Save link as..., or similar).
    Important: Do not change the name of the ".TextGrid" file, since it must match exactly the name of the .wav file. Example: "s09.objects.1.A.wav", "s09.objects.1.A.TextGrid".
  • Now you are ready to begin labeling with Praat.

3. Correcting common mistakes in the .words files

In the ortographic tier you might find some common mistakes. If you do, please correct them and submit the .words file along with the rest.

4. Guidelines and examples

• Whispery or creaky voice

  Try to label as much as possible of this, using X*?, X-? and X%? when you cannot tell the type of pitch accent or phrasal tone.

• HiF0 and pitch halving/doubling

  If the HiF0 occurs in a region that is pitch halved or doubled, mark it as usual in the tones tier and specify its actual value at the same position in the misc tier using the notation "HiF0:215".

• Restarts (1p and %r)

  (We do not use the "isofrag" label.)

• Interruptions

  If the speaker you are labeling is interrupted by the other speaker, mark so in the misc tier with the special "interrupt" label.

• Folder with all examples

5. Labeling ambiguities

• L- L% vs. !H- L%

  In cases where it is not clear how best to indicate a gradual final fall, use X-? L%. In the example, listen to the word "is".

      TextGrid file

• L* vs. !H*

  Sometimes at the end of a series of downsteps it is hard to differentiate between L* and !H*. In cases in which both pitch accents are acceptable, choose !H*. Look at the word "mime" in this example.

      TextGrid file

• Folder with all examples

6. How to save and submit your work

Once you have finished labeling a complete .wav file, follow these steps:

1. WaveSurfer: Send the new ".breaks", ".tones", and ".misc" files to agus [at] cs.columbia.edu, maintaining the original file names. If you modified the ".words" file, submit it too.
   Praat: Send the new ".TextGrid" file to agus [at] cs.columbia.edu.
2. Go to the File Locking page.
3. Click "EDIT" in the "ToBI Labeling" task of the wav file you have labeled.
4. On the next screen, change the task's status to "finished", and enter the current date.