You Described, We Archived: A Rich Audio Description Dataset

The You Described, We Archived dataset (YuWA) is a collaboration between San Francisco State University and The Smith-Kettlewell Eye Research Institute. It includes audio description (AD) data collected worldwide 2013-2022 through YouDescribe, an accessibility tool for adding audio descriptions to YouTube videos. YouDescribe, a web-based audio description tool along with an iOS viewing app, has a community of 12,000+ average annual visitors, with 3,000+ volunteer describers, and has created over 5,500+ audio described YouTube videos.

Blind and visually impaired (BVI) viewers request YouTube videos that are saved to a wishlist and volunteer audio describers select a video, write a script, record audio clips, and edit clip placement to create an audio description. The audio description tracks are stored separately and played together with the YouTube video then posted for public view at YouDescribe

The YuWA dataset covers a vast domain of videos in 15 titled categories including Film & Animation, Music, Autos & Vehicles, Travel & Events, Pets & Animals, Sports, People & Blogs, Gaming, Comedy, Entertainment, How-To & Style, News & Politics, Nonprofits & Activism, Education, Science & Technology. A video can have multiple audio descriptions and an audio description can have multiple audio clips recorded by volunteer describers. The audio clips recorded before May, 2020 were transcribed using Listen By Code and the audio clips recorded after that are transcribed using Google Cloud Speech to Text API. Viewers can rate the audio descriptions on a scale ranging from 1-5 (1 being poor, 5 being excellent). Viewers can also provide feedback to the describers by selecting some improvements from the list.

The YuWA data repository includes all YouDescribe related audio descriptions from 2013-2022 and can be sorted to include or exclude important YouDescribe milestones. We have focused on data collected by YouDescribe since March 17, 2017 and Google Analytics data which started tracking traffic since July 30, 2020. This scalable dataset will be regularly updated as new videos, audio descriptions and audio clips gets uploaded.

Run Instructions

The download_yd_data.py file was tested using Python 3.9. So, please make sure that when you use python, your Python version is at least Python 3 or make sure you specify python3.

Install the requests module:

pip install requests

# If using python3
pip3 install requests

Run the python file:

# The default configuration will store the audio clips in the current directory
# separated by YouTube video ID and Audio Description ID.
# --audioDescDir: This option allows you to specify the output directory where
#                 the audio clips will be stored.

python download_yd_data.py

# If specifying python3
python3 download_yd_data.py

# Specify output directory
python download_yd_data.py --audioDescDir=<PATH_TO_OUTPUT_DIR>

Follow the onscreen instructions when running the python file to register with the YuWA system and receive an API key to access the audio clips.
After you receive an API key which should be written in a file called yuwa.json, run the python file again and follow the on-screen instructions.