Speech Evaluation Research Study

Participant Information Sheet

What is this research about?

This research aims to investigate how well synthetic speech, generated using advanced AI techniques, replicates the characteristics of impaired speech. We evaluate these synthetic voices using human judgment.

Why are you doing this research?

Developing robust Automatic Speech Recognition (ASR) systems for people with speech impairments like dysarthria is challenging due to limited and varied real-world data. This project explores methods for augmenting such datasets with high-quality synthetic speech. These synthetic samples could make speech technologies more inclusive and accurate, especially in clinical or assistive technology settings.

Why have I been invited to take part?

You are a native English speaker who is over 18, and you have been invited to evaluate speech samples based on their naturalness and intelligibility.

What will happen if I decide to take part?

The total time required is approximately 15 to 20 minutes. You will listen to a series of short audio recordings of impaired speech (both real and AI-generated) and rate their naturalness and intelligibility. The recordings will be randomised and anonymised.

Participation is completely voluntary, and you are free to withdraw at any time without explanation or penalty.

How will the data be used?

Your responses will be stored anonymously and statistically analysed. All speech samples are pre-generated - you will not be recorded. This research is purely academic with no commercial purpose.

How will your privacy be protected?

You will be assigned an anonymous ID. No personal information will be published or disclosed. All data will be securely stored per TU Dublin's data protection policies.

Research Funding

This research is funded by Research Ireland under D-REAL (https://d-real.ie/) and ADAPT (https://www.adaptcentre.ie/).

Benefits of participation

You will contribute to advancing inclusive AI technologies for speech therapy and assistive communication.

Risks

There are minimal risks. Some speech samples may be hard to understand. You can withdraw immediately if you feel any discomfort.

Contact

For questions or concerns: D23126641@mytudublin.ie

Ethics Review

This project has been reviewed and approved by the Research Ethics Committee at TU Dublin.

Consent Form

Please confirm the following statements:

Prolific ID (Optional):

Pathological Speech Evaluation Experiment

Thank you for participating! You will now evaluate speech samples.

Instructions:

For each audio sample:

  1. Listen to the audio sample carefully
  2. Select a naturalness rating using the radio buttons (1=Bad, 2=Poor, 3=Fair, 4=Good, 5=Excellent)
  3. The transcript will automatically appear when you select your naturalness rating
  4. Select an intelligibility rating after seeing the transcript
  5. Click "Submit Rating" to save both ratings and move to the next sample

Rating Scale:

  • 1: Bad
  • 2: Poor
  • 3: Fair
  • 4: Good
  • 5: Excellent
Naturalness Rating

How natural (pleasantly human-like) was the sound of this audio sample?