In this section you will find some examples for the sounds in this task and some attributes associated with them.

1. What do we mean by a "bright" sound?
By bright we mean the sound sample is vibrant or has many high frequencies (high pitched sounds). Think of sounds made by a wooden stick hitting or scratching a metallic surface.
Example :

2. What do we mean by a "dull" sound?
We think of dull sounds as being opposite to bright. Such sounds will have more lower frequecies. Think of sounds made by a wooden stick hitting or scratching a soft cushion or a soft mattress.
Example :

3. What do we mean by a "sharp hits" or "sharp impact" sounds?
By sharp hits we mean sounds which are created when a wooden stick strikes a surface sharply. Without scraping or scratching the surface. Sharp hits can be both bright or dull.
Example :

4. What do we mean by "scratches" or "scratchy" sounds?
We think of scratchy sounds as being opposite to sharp hits. By this we mean sounds which are created when a wooden stick scratches or scrapes a surface. Scratches can be both bright or dull.
Example :

5. What do we mean by "high rate" sounds?
By Rate we mean the number of hits or scratches in a sound sample. By high rate we mean sound samples which have more than 2-3 hits or scratches.
High Rate Example :

6. What do we mean by "low rate" sounds?
By Rate we mean the number of hits or scratches in a sound sample. By low rate we mean sound samples which have less than or equal to 2-3 hits or scratches.
Low Rate Example :
Welcome to 'Qualitative evaluation of artificially synthesized audio'

Thank you for your interest in this audio evaluation task on Mechanical Turk. This page outlines the details with respect to your consent to participate in our study.


Protocol Title:
Qualitative evaluation of artificially synthesized audio


Purpose of this research:
The purpose of this research is to evaluate the quality of synthetic audio clips generated using an artificial neural network. The audio clips under test belong to the domain of audio textures (for e.g. a series of popping sounds, the ticking sound made by a Geiger counter or the sound made by water filling a container) or pitches made by musical instruments. The specific goal of this project is to ensure that the synthetic outputs generated by these deep learning models are validated by human beings to test for certain sound qualities like smoothness, naturalness or absence of noise etc.


Contacting Us:
Please use the "Contact this requester" link on the Mechanical Turk's HIT details section if you need further details or clarifications on the task. The Principal Investigator (the person in charge of this research) or his/her representative will answer all of your questions. Note: On using this feature, your e-mail address and your name will automatically be visible to our research team so that we can reply to you.


Your personal data:
Only the principal investigator and his/her representative has your worker ID and this will not be released to any other person. We will NOT request Amazon Mechanical Turk for your person identifying details. Also, this worker ID data will not be used in a publication or presentation. We will use this worker ID only for the purposes of correlating consent to HIT responses and distributing compensation via Mechanical Turk.

All data collected will be kept in accordance to the University's Research Data Management Policy. Research data used in any publication will be kept for a minimum of 10 years before being discarded.


By accepting the HIT on Amazon's Mechanical Turk, you acknowledge that -
  1. You have agreed to take part in this research.
  2. You have read the instructions on the mechanical turk website and details on this form which explains the use of the data you submit for this research.
  3. You can withdraw from this research at any point of time. If you want the data you already submitted to be deleted you will inform the Principal Investigator and all your data will be discarded.
  4. You will not have any financial benefits that result from the commercial development of this research.
  5. You agree to be re-contacted for future related studies via Amazon's Mechanical Turk website.  
 

Note: Please use headphones to listen to the audio samples. Some samples may not be audible over laptop/desktop speakers.

 
Audio Evaluation Task
 
 
In this task, you will listen to some sounds made by a wooden stick hitting various different hard and soft surfaces and compare them to each other.
 
See some examples sounds and attributes associated with them, please see the 'View Instructions' section.
 
Please listen to the two audio samples and answer the questions below.
Sample 1:
 
Sample 2:
 
1. How does Sample 1 differ from the Sample 2 in terms of brightness?

By bright we mean the sound sample is vibrant or has many high frequencies (As compared to dull sounds).


See 'View Instructions' for examples.

 


 

 

2. How does Sample 1 differ from the Sample 2 in terms of rate?

By high rate we mean the sound samples have more than 2-3 hits or scratches (As compared to low rate).

See 'View Instructions' for examples.

 


 

 

3. How does Sample 1 differ from the Sample 2 in terms of impact type (hit or a scratch)?

By scratches we mean sounds which are created when a wooden stick scratches or scrapes a surface (As compared to sharp hits).

See 'View Instructions' for examples.