한국어 발화에서 발음변이 추적을 위한 발음열 변환
Grapheme to Phoneme Conversion for Tracing Pronunciation Deviation
1
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
The Korean writing system is supposed to be based on morphophonemic representations. This means that written symbols can be matched to actual pronounced segments after application of phonological rules that map morphophonemic to phonemic representations. Various grapheme-to-phoneme (GTP) conversion systems have been proposed for the purpose of generating pronunciation strings for speech synthesis as well as a pronunciation dictionary for automatic speech recognition. These conversion systems were developed to generate a pronunciation string that could be expected if someone pronounces a given word string. However, there are some words whose pronunciations are deviant from the expected ones. In order to investigate how different pronunciations are in real speech, we need to develop a GTP conversion program that can not only produce expected pronunciations but also indicate which phonological rules are applied, so that we can trace the source of the variation. In this paper, we explain how we developed a GTP conversion system that can generate the expected pronunciations of given words and the phonological rules that are applied to the word string.
Abstract
The Korean writing system is supposed to be based on morphophonemic representations. This means that written symbols can be matched to actual pronounced segments after application of phonological rules that map morphophonemic to phonemic representations. Various grapheme-to-phoneme (GTP) conversion systems have been proposed for the purpose of generating pronunciation strings for speech synthesis as well as a pronunciation dictionary for automatic speech recognition. These conversion systems were developed to generate a pronunciation string that could be expected if someone pronounces a given word string. However, there are some words whose pronunciations are deviant from the expected ones. In order to investigate how different pronunciations are in real speech, we need to develop a GTP conversion program that can not only produce expected pronunciations but also indicate which phonological rules are applied, so that we can trace the source of the variation. In this paper, we explain how we developed a GTP conversion system that can generate the expected pronunciations of given words and the phonological rules that are applied to the word string.
참고문헌 (17)
- 국립국어연구원, "한국어문규정집", 국립국어연구원, ~ (2011)
- 김선희, "한국어 자동 발음열 생성을 위한 예외발음사전", 음성과학, 10(4), pp.167~180 (2003), 한국음성과학회
- 김선희, "최소 형태소 정보를 이용한 자동 발음열 생성 시스템", 2003년 한국멀티미디어학회 추계학술발표대회논문집, pp.216~219 (2003)
- 김흥규, "21세기 세종계획 국어 기초자료 구축" (2007), 국립국어원
- 안병섭, "음운론적 수의성의 개념 정립을 위한 검토", 동악어문학, (61), pp.145~176 (2013), 동악어문학회
- 이경림, "한국어 연속음성 인식을 위한 발음열 자동 생성", 한국음향학회지, 20(2), pp.35~43 (2001)
- 이창헌, "실시간 연속 대용량 한국어 음성 인식기의 개발 및 자동발음열 생성", 2014년도 대한전자공학회 하계학술대회, 37(1), pp.643~644 (2014)
- 이호영, "국어음성학" (1996), 태학사
- 정경석, "한글 TTS 시스템에서 문자열-발음열 변환기의 ㄴ-소리 첨가 전 처리기", 2001년도 한국정보과학회 봄 학술발표논문집, 28(1), pp.370~372 (2001)
- 차선화, "TTS 시스템을 위한 한국어 발음열 자동 생성", 제15회 음성통신 및 신호처리 워크샵, 15(1), pp.414~418 (1998)
- 홍성훈, "음운론의 계량적 방법론" (2014), 한국문화사
- Garofolo, J. S., "TIMIT Acoustic-Phonetic Continuous Speech Corpus" (1993), Linguistic Data Consortium
- Jun, S. A., "The Accentual Phrase in the Korean Prosodic Hierarchy", Phonology, 15(2), pp.189~226 (1998)
- Lee, J., "A Data-driven Grapheme-to-Phoneme Conversion Method Using Dynamic Contextual Converting Rules for Korean TTS System", Computer Speech and Language, 23, pp.423~434 (2009)
- Pitt, M. A., "The Buckeye Corpus of Conversational Speech: Labeling Conventions and a Test of Transcriber Reliability", Speech Communication, 45, pp.89~95 (2005)
- Yoon, K., "A Linguistically Motivated Approach to Grapheme-to- Phoneme Conversion for Korean", Computer Speech and Language, 20, pp.357~381 (2006)
- 윤원희, "The Korean Corpus of Spontaneous Speech", 말소리와 음성과학, 7(2), pp.103~109 (2015), 한국음성학회[DOI]