A Comparative Study on English and Korean Vowel Formants


The purpose of this study is to analyze read speech corpus of previous studies and compare that data to spontaneous speech corpus by examining the vowel properties between read and spontaneous speech in English and in Korean.
A significant number of vowel formant studies utilize read speech corpus. The research procedures are easy to employ and the linguistic factors are easy to control. Although researcher and participant bias may influence the outcome with artificial recordings, these conditions are advantageous for data extraction. Compared to read speech, spontaneous speech data requires a more complex analysis because the procedures and the linguistic factors influence the amount and type of data output. Spontaneous speech produces a considerable amount of unneeded data and controlling linguistic factors can be difficult. However, recorded spontaneous speech produces more authentic output in spite of researcher bias.
The previous research studies provided data for analysis of vowel formants in both English and Korean read speech. In both languages, read speech data revealed wider vowel spaces, narrower degree of scattering among vowel data and differences in locations of vowel spaces by sex.
The Buckeye corpus and the Seoul corpus were used to examine vowel formants of spontaneous speech in English and Korean. Analysis of vowel properties in the English spontaneous speech revealed that the vowel formant distribution was affected by different linguistic factors that included the following: sex, age, speech rate, word frequency, syllable stress of target vowel, vowel location in a word or utterance and content versus function word. The linguistic factors affecting vowel properties in Korean spontaneous speech included: sex, age, speech rate, interviewer sex, target vowel location in a phrasal word or utterance and location of syllable containing the target vowel with respect to a phrasal word or utterance. These variables from both languages diversly affected the vowel formant distribution. The results revealed that a similarity exists between the formant distribution of the spontaneous speech and the read speech. In addition, the spontaneous speech at times presented irregular aspects and unique properties in the formant distribution. The data from this research brings forth new understanding of vowel properties relative to read and spontaneous speech in English and Korean.