Voice Data Labeling System

• This project is currently being used in the research center of the University of Pécs

Aim of the project

To have a system to support quick, easy, and user-friendly labelling of voice data. In supervised machine learning projects, it is crucial to have a good database for training, validation and test of the ML model.

Feautures

• The application receives the root folder and show folder structure in a TreeView
• Provide an option to select a source folder and start labelling in this selected folder
• List command types stored in a file
• List command (labels) of selected command type in alphabetic order
• On folder selection in in folder structure, list sample files in alphabetic order
• Play selected sample file (wav, mp3)
• Rename folder to selected command type
• Rename sample file to selected command
• Mark a sample as useless
• After confirmation, automatically select next command and next sample
• List of useless samples is written into a log file