A corpus-based approach to dialectal variation in Korean vowels

This paper provides an acoustic study of monophthongal vowels (i e ɛ ɨ ʌ a u o) in eight South Korean dialects based on a dialectal speech corpus created by the National Institute of the Korean Language (2004-2011). The spontaneous speech data were automatically segmented by force-aligning them with phonemic transcriptions. Multiple formant values using different LPC formant ceiling settings were extracted and the formant values that maximize the speaker’s vowel space, as calculated by Heron’s method, were selected as optimal. The resulting vowel space of the eight dialects, extracted using minimal manual correction, shows patterns that are largely consistent with the traditional impressionistic descriptions. The study lays a foundation for further acoustic exploration of the spontaneous dialectal speech using computational tools. The study makes empirical and methodological contributions to the study of Korean dialects and corpus-based dialect study more generally.