create new data sets for experiments

• preparation:

  1. copy cerevoice_eng/ and voice/ folders into preprocessing/
     • voice/ has to contain both voice and licence files
  2. optional: copy bbdf and refdf files to preprocessing/data/

• usage: run preprocessing/create_json.py with python 2.7 (--h for list of available arguments)

  • specify location of bbdf and refdf files with --o , if files are not in preprocessing/data
  • default output folder: preprocessing/data/output

evaluate log files

• preparation:

  1. copy images/ , audio/ and json/ from Slurk static folder into evaluation/data/
  2. copy corresponding log file(s) into evaluation/data/

• usage: run evaluation/evaluate_logs.ipynb with Jupyter Notebook

AMT integration

• preparation

  1. create config.ini (copy from config.template.ini)
  2. fill in AWS credentials, change urls if slurk is not hosted locally ('link_generator' and 'login' sections)
  3. set 'value' to 'true' if used in production ('false' for sandbox)

• publishing HITs

  • execute publish_hits.py to publish HITs on AMT (specify number of HITs using the -n argument)
  • all HIT IDs are registered in JSON files in the 'published' directory

• deleting HITs

  • execute delete_hits.py to delete all HITs found in the 'published' directory