Speech exemplar and evaluation database (SEED) for clinical training in articulatory phonetics and speech science

ABSTRACT

One challenge faced by teachers of phonetics, speech science, and clinical speech disorders courses is providing meaningful instruction that closes the theory to practice gap. One barrier to providing this type of deep learning experience is the lack of publicly available examples of speech recordings that illustrate comparisons between typical and disordered speech production across a broad range of disorder populations. Data of this type exist, but are typically collected for specific research projects under narrowly written IRB protocols that do not allow for release of even de-identified speech recordings to other investigators or teachers. As a partial corrective to this problem, we have developed an approved publicly available database of speech recordings that provides illustrative examples of adult and child speech production from individuals with and without speech disorders. The recorded speech materials were designed to illustrate important clinical concepts, and the recordings were collected under controlled conditions using high-quality equipment. The ultimate goal of creating this corpus is to improve practitioners’ and scientists’ understanding of the scientific bases of knowledge in our profession and improve our ability to develop clinical scientists and young researchers in the field.

KEYWORDS:

• Database
• speech production
• child speech
• adult speech
• speech disorders

Acknowledgments

This database was funded by a University of Cincinnati Faculty Development Award to Suzanne Boyce and the Auburn University Breeden Endowment Award for Teaching Innovation to Marisha Speights and Dallin Bailey. Collection of child speech samples was also funded by the Auburn University Office of the Vice President for Research and Economic Development Internal Grant Program Innovation Award. Katherine Willoughby offered volunteer hours as an undergraduate and graduate student to organize and maintain the data management system. We are most grateful for all the undergraduate and graduate research assistants at the University of Cincinnati and Auburn University who aided in the segmentation of the speech samples. Student work was supported through the Breeden Endowment Award and the Auburn University College of Liberal Arts Stevens Research Fund. Without our students and financial support, this contribution would not be possible.

Disclosure of interest

The authors report no conflict of interest.

Additional information

Funding

This work was supported by the Auburn University [Breeden Endowment Award for Teaching Innovation, College of Liberal Arts Stevens Research Fund, Office of the Vice President for Research and Econ]; University of Cincinnati [Faculty Development Award].