LangDiv: HNDS-R: Talker-specific, language-specific, and language-general characteristics of first-language and second-language speech

In today’s globalized societies, many conversations involve second-language speakers whose pronunciations typically exhibit a noticeable foreign accent. This project asks how foreign-accented speech varies across talkers from different first-language backgrounds and whether there are common features that characterize speech as foreign-accented across various languages.

A detailed description of foreign-accented, or second-language, speech across individuals and across languages helps to develop strategies for overcoming miscommunications that can hinder conversations between humans from different language backgrounds and limit the effectiveness of human-computer voice interactions. This project also supports hands-on education for language sciences research.

This project has two specific aims. The first aim develops a unique resource of digital speech recordings by both first-language and second-language speakers of multiple languages. Each speaker is recorded reading a carefully selected set of sentences and in both their first-language and their second-language.

The recordings also include casual, spontaneous speech by each speaker in both of their languages. This unique set of recordings is fully documented, annotated, and made publicly available to encourage its use by researchers, speech technology developers, educators, and clinicians. In the second aim, the research potential of the digital speech resource is demonstrated through a series of acoustic analyses that develop a detailed and comprehensive acoustic characterization of first-language versus second-language speech across talkers and across languages.

The analyses extend prior studies of variation in first-language and second-language speech that were challenged by the lack of parallel recordings in multiple languages. Collectively, this project yields a novel, publicly available digital speech resource and a set of analyses that can be leveraged for research and broader theorizing on both human and computer speech communication in multi-lingual settings.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.