AWS Mechanical Turk Audio Classification
NYU Research
This is a webpage made in order to collect data on audio samples for training AI A Worker is presented an audio sample, and then asked whether or not a certain something is present in the audio. Afterwards the user is either asked:
- Their answer confidence
- Whether the user would pick a 50:50 lottery or their answer's correctness for a chance to win bonus pay.
- If the user picks lottery, the chances are raised (e.g. 60:40) and are asked again
- If the user picks their answer or the lottery has reached 100:0, the user proceeds to the next question
nitty-gritty:
Two parameters must be passed in addition to the amazon parameters in Mechanical Turk (mturk).
task_type = 1 || 2
items = [
{
'label': 'Thing that might be in the audio',
'file': 'The file located in train_audio.wav'
}
]
They must be added when the Hit is created. The Python3 script post_hits.py shows how this is done.
Note, you will have to set your own environment variables to use boto3.Session()
I personally set AWS_SHARED_CREDENTIALS_FILE=C:\someroute\to\credentials
It is currently set up so mturk points directly to this github site.
Usual setup
Project setup
npm install
Compiles and hot-reloads for development
npm run serve
Compiles and minifies for production
npm run build
Run your tests
npm run test
Lints and fixes files
npm run lint