audio-annotation

A web interface for annotating audio data.

Description

audio-annotation is a web application that allows users to annotatate audio files. It is developed using Python (Flask), Javascript, HTML5/CSS3, and Postgresql. It is extended from audio-annotator. The annotation progress is splitted into two major steps: Coarse-grained and Fine-grained. In Coarse-grained, the annotator should only indicate the presence of a desired label. The audio files that have the label, will be moved to Fine-grained annotation for a deaper labeling potentially with different passes for one or more passes.

1. Coarse-grained step

2. Fine-grained step

Benefits:

1. Admin page to view and manage tables.
2. Definable labels and annotation passes
3. Can measure annotator agreements by letting each segment be annotated with more than one person

Data format

WORK_PATH= is the starting directory. ORIGINAL_FILES_DIRECTORY= is the audio file directory (in this directory the folders should be named based on the participant and their audio recordings are located inside them) SEGMENT_FILES_DIRECTORY= After removing silence intervals and performing segmentation, the audio segments will be automatically created here.

Database setup

1. Download postgresql.
2. createdb databasename

environmental variables:

• export EMAIL_SENDER= The email used for sending errors
• export EMAIL_PASSWORD= The email password
• export EMAIL_RECEIVER= The email used for receiving errors
• export DB_NAME= Database name
• export DB_USER= Database username
• export DB_PASSWORD= Database password
• export FIRST_RUN=True # for the first time running the server, otherwise False

Run app

gunicorn --bind 0.0.0.0:5000 --workers=1 --timeout=0 --access-logfile access.log run:app