Creating a diarization broadcast corpus

Requirements

• Gecko or Audacity,
• rttm files (if you have linux then you can generate the rttm files yourselves)
• corresponding videos/audio of episode

Tips

Label speaker turns which last at least 90 ms. (CHANGED)

Each speaker gets their own speaker number per recording.

Add all the speakers to one segment and copy over the list then remove them back all again to create initial list for the csv file.

Unknown speakers get labelled Unknown 01 etc.

Process

1. Generate the proposed rttm files for 28 episodes that week.
2. Labelling - Gecko
   1. Open Gecko If you use the Gecko version linked here then you can save partially corrected files and reload them back into the editor.
   2. Upload the video file & rttm file
   3. Adjust the segment start and end times to match speaker turns.
   4. Add missing speaker turns.
   5. Correct speaker labels/numbers. Add new ones if necessary
   6. Write down the speaker names which correspond to each speaker number. These go in a csv file.
   7. Label music, foreign language, or noise. They're available as default labels.
   8. Export as json, srt, and rttm.
3. Labelling - Audacity
   1. Open Audacity
   2. Upload the video file & label file
   3. Adjust the segment start and end times to match speaker turns.
   4. Add missing speaker turns.
   5. Correct speaker labels/numbers. Add new ones if necessary
   6. Write down the speaker names which correspond to each speaker number. These go in a csv file.
   7. Label music, foreign, or noise in square brackets.
   8. Export label file.
4. Repeat for a new episode.
5. Turn in the corrected rttm, json, srt, and csv files then get new rttm and video files.
6. Judy reports the new DER with that week's data.

reco2spk_num2spk_name

format

<recording/episode id>, <speaker_number in rttm file>, <speaker name>

example

Fréttirkl1900-5022010T0,1, Bogi Águstsson