Current Practices in Voice Data Collection and Limitations to Voice AI Research: A National Survey

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Speech Science

Background:

• Voice AI algorithm accuracy depends on high-quality voice data.
• Existing voice data is underutilized due to a lack of standardized acoustic data management protocols.
• This limits the potential for large-scale voice artificial intelligence (AI) research.

Purpose of the Study:

• To assess current practices in voice data collection, storage, and analysis within North American voice centers.
• To identify perceived limitations hindering collaborative voice research.
• To inform the development of standardized protocols for voice data management.

Main Methods:

• A 30-question online survey was distributed to practitioners at North American voice centers.
• The survey collected data on acoustic data management practices and barriers to collaboration.
• Respondents included laryngologists and speech-language pathologists (SLPs).

Main Results:

• Only 28% of respondents use standardized protocols for acoustic data collection and storage.
• Despite 87% conducting voice research, only 38% collaborate across institutions.
• Key limitations include lack of standardized methodology (30%) and insufficient resources for data preparation (55%).

Conclusions:

• There is a significant need for standardized acoustic data management to enable large-scale, multi-institutional voice research using AI.
• Developing infrastructure for secure and efficient data sharing is essential.
• Standardization will improve the usability and validity of voice datasets for AI development.